conference theme: better decisions for africa

102
41 st Annual Conference of the Operations Research Society of South Africa Conference Theme: Better Decisions for Africa — Programme — Aloe Ridge Hotel and Game Reserve, Muldersdrift 16–19 September 2012

Upload: others

Post on 26-Oct-2021

7 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Conference Theme: Better Decisions for Africa

41st Annual Conference of the

Operations Research Society of South Africa

Conference Theme: Better Decisions for Africa

— Programme —

Aloe Ridge Hotel and Game Reserve, Muldersdrift16–19 September 2012

Page 2: Conference Theme: Better Decisions for Africa

Welcome from the President

It is with great pleasure that I welcome all delegates to the41st Annual Conference of the Operations Research Society ofSouth Africa (ORSSA) — the highlight in our Society’s calendaryear, this year held at the beautiful Aloe Ridge Hotel and GameReserve.

The conference was organised by the Pretoria Chapter of ORSSA,assisted by the Johannesburg Chapter. I would like to thankWinnie Pelser, chair of both the Pretoria Chapter and the Local

Organising Committee (LOC) of the conference, as well as Dave Evans, LouisDannhauser, Elias Willemse, Marthi Harmse and all the other members of the LOCfor the hard work they have put in on many fronts and in various capacities overmany months in order to bring this meeting here to the Cradle of Humankind.

Our conference programme this year boasts an impressive array of diverse presenta-tions on the development of new theory, on the application of operational researchtechniques in business and industry, on topical issues in operations research, andon the philosophy, teaching and marketing of operations research. This rich pro-gramme, comprising a total of 73 papers, a tutorial and an advertorial, has beenorganised into three plenary sessions and twenty seven contributed sessions, runningin three or four parallel streams for virtually our entire stay at the Aloe Ridge Hotel.This exciting scientific programme promises to cater for every delegate, no matterwhat their particular tastes and preferences!

We welcome, in particular, our keynote speaker, Professor John Hearne (RoyalMelbourne Institute of Technology, Australia), who needs little introduction atORSSA and who will be delivering both the opening and closing plenary lecturesof the conference on the fascinating topics of wildlife land management and theinterplay between conservation and the commercial utilisation of our wildlife.

In addition to interesting and engaging scientific content, I hasten to mention thatORSSA conferences have of course also been famous for decades now for theirthoroughly enjoyable social aspects of the conference programme, and this confer-ence promises to be no exception! I trust that all delegates will have a productive fewdays here at the Aloe Ridge Hotel, exchanging valuable and inspiring ideas, learn-ing new tricks of our diverse and complex trade, renewing old acquaintances, andmaking new friends and colleagues. May you enjoy the conference in the beautifulsurroundings of the Aloe Ride Game Reserve.

Best wishes,

Jan van Vuuren, PresidentOperations Research Society of South Africa

ii

Page 3: Conference Theme: Better Decisions for Africa

Welcome from the Chair of the Organising Committee

Welcome to the 41st Annual Conference of ORSSA. We are proudto welcome you to Aloe Ridge Hotel and Game Reserve in closeproximity to the Cradle of Humankind. We hope you will enjoythe beautiful and interesting surroundings.

We are pleased to have Professor John Hearne (Royal MelbourneInstitute of Technology, Australia) as the keynote speaker. Profes-sor Hearne is a familiar face in ORSSA. He is a past president ofORSSA and a multiple recipient of the Tom Rozwadowski award.

In view of ORSSA’s outreach and greater collaboration across operational researchcommunities in Southern Africa and the very successful conference in Zimbabwelast year, the theme for the 2012 conference is: Better decisions for Africa. Weare therefore very pleased to have delegates from within South Africa and fromZimbabwe, Uganda and Cameroon.

The programme contains more than 70 interesting and diverse papers ranging fromtheory to applications in true Operations Research tradition. We trust that you willhave plenty of opportunities to exchange knowledge, experience and new develop-ments, and very importantly, renew old acquaintances and make new friends.

May you enjoy this 2012 conference of ORSSA in the tranquil South African bush!

Best wishes,

Winnie Pelser, ChairORSSA 2012 Local Organising Committee

Members of the ORSSA 2012 Local Organising Committee:

Winnie Pelser, Chair (Arms Corporation of South Africa)Louis Dannhauser, Vice Chair (Vodacom (Pty) Ltd)Wilna Bean (Council for Scientific and Industrial Research)Brahm Bothma (RTT Medical)Ian Campbell (University of the Witwatersrand)John Dean (Private Capacity)Dave Evans (Development Bank of Southern Africa)Elias Munapo (University of South Africa)Derek Saunderson (Private Capacity)Nadia Viljoen (Council for Scientific and Industrial Research)Elias Willemse (LTS Consulting)

iii

Page 4: Conference Theme: Better Decisions for Africa

— Table of Contents —

Overview Map of the Aloe Ridge Hotel Layout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1

Detailed Map of Main Hotel Area and Conference Facilities . . . . . . . . . . . . . . . . . . . . . 2

Conference Programme at a Glance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

Detailed Conference Programme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

List of Sessions & Chairs (in chronological order) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

List of Paper Titles (in alphabetical order) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

List of Authors (in alphabetical order) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .18

Abstracts of Plenary Papers (in chronological order) . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

Abstracts of Contributed Papers (in alphabetical order) . . . . . . . . . . . . . . . . . . . . . . . 24

List of Delegates (in alphabetical order) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

Sponsors & Service Providers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

— Notes —

iv

Page 5: Conference Theme: Better Decisions for Africa

— Overview Map of the Hotel —

Gorge

Room

WaterfallBar

Observatory

Restaurant

Hedwig’sRestaurant

TennisCou

rts

ThePlatform

Outdoor

Swim

mingPool

GuestRoom

s

GuestRoom

s

MainHotel

Area

ParkingLot

Entran

ceMain

1

Page 6: Conference Theme: Better Decisions for Africa

— Map of Conference Facilities—

Gorge

Room

MainDiningRoom

Bar

Indoor

Swim

mingPool

MainEntran

ce

Jacaran

daRoom

Reception

Stairs

AloeRoom

Lift

WC

Coff

ee/T

eaStation

SquashCou

rt&

Gam

esCentre

RegistrationDesk

Orchid

Room

ParkingLot

2

Page 7: Conference Theme: Better Decisions for Africa

— Programme at a Glance —

Mon

day

17

Sep

tem

ber

2012

08:0

0–08

:30

On

site

Reg

istr

ati

on

(Fo

yer

ou

tsid

eth

eA

loe

Roo

m)

08:3

0–10

:00

Plenary

Session

A:OpeningKeynote

Addre

ssbyPro

fessorJohn

Hearn

e(A

loe

Room

)10

:00–

10:3

0T

ea/C

off

ee(N

ext

toth

eIn

doo

rP

ool)

10:3

0–12

:30

I:M

isce

llan

eou

sT

opic

sII

:H

euri

stic

s&

Met

a-

III:

Mis

cell

an

eou

sT

op

ics

IV:

Tact

ical

Sit

uati

on

(Alo

eR

oom

)h

euri

stic

s(G

org

eR

oom

)(J

aca

ran

da

Room

)A

ware

nes

s(O

rch

idR

oom

)12

:30–

13:4

5L

un

ch(M

ain

Din

ing

Roo

m)

13:4

5–15

:15

V:

Sim

ula

tion

VI:

Com

bin

ato

rial

VII

:T

ech

nolo

gy

&E

du

cati

on

VII

I:C

ase

Stu

dy

Tu

tori

al

(Alo

eR

oom

)O

pti

mis

ati

on

(Gorg

eR

oom

)(J

aca

ran

da

Room

)(O

rch

idR

oom

)15

:15–

15:4

5T

ea/C

off

ee(N

ext

toth

eIn

doo

rP

ool)

15:4

5–17

:15

IX:

Th

eP

etro

-ch

emic

alX

:M

etah

euri

stic

s&

AI

XI:

Net

work

s&

Con

nec

tivit

yX

II:

OR

SS

AE

xec

uti

ve

Ind

ust

ry(A

loe

Room

)(G

org

eR

oom

)(J

aca

ran

da

Room

)M

eeti

ng

(Orc

hid

Room

)19

:00–

Barb

ecu

e(T

he

Pla

tform

,w

eath

erpe

rmit

tin

g)

Tu

esd

ay

18

Sep

tem

ber

2012

08:3

0–09

:50

Plenary

Session

B:Advertorialand

Mid-confere

nceKeynote

Addre

ss(A

loe

Room

)10

:00–

11:0

0X

III:

An

alyti

csX

IV:

Veh

icle

Rou

tin

gX

V:

Dec

isio

nM

akin

g(A

loe

Room

)(J

aca

ran

da

Room

)(O

rch

idR

oom

)11

:00–

11:3

0T

ea/C

off

ee(N

ext

toth

eIn

doo

rP

ool)

11:3

0–12

:30

XV

I:In

vest

men

tP

ortf

olio

An

alysi

sX

VII

:A

gri

cult

ure

XV

III:

Mu

lti-

crit

eria

Dec

isio

n(A

loe

Room

)(J

aca

ran

da

Room

)A

naly

sis

(Orc

hid

Room

)12

:30–

13:4

5L

un

ch(M

ain

Din

ing

Roo

m)

13:4

5–15

:15

XIX

:O

pti

mis

atio

nin

Ind

ust

ryX

X:

Pre

dic

tive

Mod

elli

ng

XX

I:H

ealt

hC

are

(Alo

eR

oom

)(J

aca

ran

da

Room

)(O

rch

idR

oom

)15

:15–

15:4

5T

ea/C

off

ee(N

ext

toth

eIn

doo

rP

ool)

15:4

5–17

:15

AnnualGenera

lM

eeting

(Alo

eR

oom

)19

:00–

Con

fere

nce

Ban

quet

(Gorg

eR

oom

)

Wed

nesd

ay

19

Sep

tem

ber

2012

08:3

0–10

:00

XX

II:

OR

inD

evel

opm

ent

XX

III:

Sch

edu

lin

gX

XIV

:S

tati

stic

al

&D

ata

An

aly

sis

(Alo

eR

oom

)(J

aca

ran

da

Room

)(O

rch

idR

oom

)10

:00–

10:3

0T

ea/C

off

ee(N

ext

toth

eIn

doo

rP

ool)

10:3

0–11

:30

XX

V:

Nat

ura

lR

esou

rce

Man

agem

ent

XX

VI:

Info

rmati

on

Tec

hnolo

gy

XX

VII

:T

ran

sport

&T

ran

sport

ati

on

(Alo

eR

oom

)(J

aca

ran

da

Room

)(O

rch

idR

oom

)11

:40–

13:0

0Plenary

Session

C:ClosingKeynote

Addre

ssbyPro

fessorJohn

Hearn

e(A

loe

Room

)13

:00–

14:0

0L

un

ch(M

ain

Din

ing

Roo

m)

3

Page 8: Conference Theme: Better Decisions for Africa

— Detailed Conference Programme —

Sunday 16 September 2012

On-site Registration (14:00–16:00)[Foyer outside the Aloe Room]

Welcome Reception (16:00–18:30)[Gorge Room]

Monday 17 September 2012

On-site Registration (08:00–08:30)[Foyer outside the Aloe Room]

Monday 17 September 2012: (08:30–10:00)

Plenary Session A: Conference OpeningChair: Dave Evans [Venue: Aloe Room]

08:30–08:45 Winnie Pelser (Chair, Local Organising Committee)Welcome & Announcements

08:45–09:00 Jan van Vuuren (ORSSA President)Presidential Address

09:00–10:00 John Hearne (Keynote Speaker)Spatial problems in managing land for habitat and fire (p. 21)

Tea/Coffee (10:00–10:30) [Next to the Indoor Swimming Pool]

Monday 17 September 2012: (10:30–12:30)

Parallel Session I: Miscellaneous TopicsChair: Hans Ittmann [Venue: Aloe Room]

10:30–11:00 Ian Campbell, Airline taxi scheduling (p. 24)11:00–11:30 Kanshukan Rajaratnam, Consumer loan decisions with a profit-

loss trade-off under multiple economic conditions (p. 40)11:30–12:00 Justin Chirima, Absalom Jaison & Desmond Mwembe, An

optimal distribution strategy for a supply chain organisation: Thecase of Delta Beverages (p. 70)

12:00–12:30 Tichoana Mazuru, Strategic coordination in a production supplychain (p. 84)

4

Page 9: Conference Theme: Better Decisions for Africa

Parallel Session II: Heuristics and MetaheuristicsChair: Elias Willemse [Venue: Gorge Room]

10:30–11:00 Francois Fagan, An articulated qualitative model for evolutionaryalgorithms (p. 29)

11:00–11:30 Bernard Schlunz, Application of a harmony search algorithm tothe core fuel reload optimisation problem for the SAFARI-1 nuclearresearch reactor (p. 26)

11:30–12:00 Michael Olusanya, Studies in metaheuristics for the BloodAssignment Problem (p. 86)

12:00–12:30 Aderemi Adewumi, A model and results of local search heuristicsfor the Hostel Space Allocation Problem (p. 64)

Parallel Session III: Miscellaneous TopicsChair: Elias Munapo [Venue: Jacaranda Room]

10:30–11:00 Anton de Villiers, An algorithm for efficient secure networkdomination (p. 25)

11:00–11:30 Moonyoung Yoon, Finding the best pass-receiving position in theRoboCup Small-Size League (p. 51)

11:30–12:00 Annette van der Merwe, Applying expert system technology indietetics (p. 28)

12:00–12:30 Jan Kruger, The best posterior probability structure is notnecessarily the causal structure (p. 33)

Parallel Session IV: Tactical Situation Awareness Decision SupportChair: Winnie Pelser [Venue: Orchid Room]

10:30–11:00 Jacques du Toit, Coastal threat evaluation decision support (p. 37)11:00–11:30 Andries Heyns, The optimisation of terrain modelling and line-

of-sight techniques (p. 71)11:30–12:00 Daniel Lotter, Design of a generic weapon assignment system in

a ground-based air defence environment (p. 42)12:00–12:30 Michelle van der Merwe, The weapon assignment scheduling

problem in a ground-based air defence environment (p. 94)

Lunch (12:30–13:45) [Main Dining Room]

Monday 17 September 2012: (13:45–15:15)

Parallel Session V: SimulationChair: Louis Dannhauser [Venue: Aloe Room]

13:45–14:15 Anette van der Merwe, Improved techniques to model continuousoperations with discrete-event simulation (p. 55)

14:15–14:45 Cecile Bezuidenhoudt, Application of optimisation methods tosimulation models (p. 27)

14:45–15:15 Mark Einhorn, Self-organising traffic light control (p. 80)

5

Page 10: Conference Theme: Better Decisions for Africa

Parallel Session VI: Combinatorial OptimisationChair: Margarete Bester [Venue: Gorge Room]

13:45–14:15 Tjaart Steyn, Empirical results: Exact solutions using limitedpattern generation for the N-sheet Cutting Stock Problem (p. 48)

14:15–14:45 Berndt Lindner, Development and application of an assignmentproblem to reduce overcut waste in a secondary wood manufacturingfacility (p. 45)

14:45–15:15 Elias Munapo, Solving a binary linear programming model inpolynomial time (p. 81)

Parallel Session VII: Technology and EducationChair: Daniel Lotter [Venue: Jacaranda Room]

13:45–14:15 Wim Gevers, Spreadsheet solvers: A comparison (p. 82)14:15–14:45 Olatunde Osiyemi, Integrating technology into teaching and

learning in the school classroom (p. 57)14:45–15:15 Olatunde Osiyemi, Teachers’ and learners’ perception on math-

ematics literacy as subject stream (p. 90)

Parallel Session VIII: Case Study TutorialPresenter: Nadia Viljoen [Venue: Orchid Room]

13:45–15:15 Using operations research for strategic planning in the not-for-profithealth sector — A case study (p. 92)

Tea/Coffee (15:15–15:45) [Next to the Indoor Swimming Pool]

Monday 17 September 2012: (15:45–17:15)

Parallel Session IX: The Petro-chemical IndustryChair: Patrick Veldhuizen [Venue: Aloe Room]

15:45–16:15 Ester Vermaak, Stochastic simulation for Sasol Solvents GlobalTank Planning (p. 83)

16:15–16:45 Nicaleen van der Westhuizen, A successful case study forintegration between business and analytics in the Propylene valuechain (p. 87)

16:45–17:15 Jaco Joubert, A practical application of stochastic simulationwithin an explosives supply chain environment (p. 74)

Parallel Session X: Metaheuristics and Artificial IntelligenceChair: Aderemi Adewumi [Venue: Gorge Room]

15:45–16:15 Benson Baha, An artificial neural network for detecting risk ofType 2 Diabetes (p. 31)

16:15–16:45 Nachamanda Blamah, An intelligent particle swarm optimisa-tion framework based on a multi-agent system (p. 58)

16:45–17:15 Sandile Saul, An artificial bees colony algorithm for the TravellingTournament Problem (p. 30)

6

Page 11: Conference Theme: Better Decisions for Africa

Parallel Session XI: Networks and ConnectivityChair: Tjaart Steyn [Venue: Jacaranda Room]

15:45–16:15 Andre Brand, A linear response surface analysis approach toevaluate QoS in wireless networks (p. 61)

16:15–16:45 Stephanus Terblanche, A problem reduction approach for theSurvivable Network Design Problem (p. 75)

16:45–17:15 Thandulwazi Magadla, The effect of network structure andswitching costs on innovation diffusion in social networks (p. 47)

Parallel Session XII: ORSSA Executive Committee Meeting(15:45–17:15) Chair: Jan van Vuuren [Venue: Orchid Room]

Barbecue (19:00–) [The Platform, weather permitting]

Tuesday 18 September 2012

Tuesday 18 September 2012: (08:30–09:50)

Plenary Session B: Mid-conference PlenaryChair: Paul Fatti [Venue: Aloe Room]

08:30–09:00 Clemens Dempers (Blue Stallion Technologies ) How to save timeand effort: Selected local case studies using the modelling and visu-alization tools of Blue Stallion Technologies (p. 34)

09:00–09:50 Patrick Veldhuizen (Sasol Shared Services, A Division of SGS)The Sasol group energy linear programming model (p. 22)

Tuesday 18 September 2012: (10:00–11:00)

Parallel Session XIII: AnalyticsChair: Caston Sigauke [Venue: Aloe Room]

10:00–10:30 Hans Ittmann, The hype around Analytics (p. 53)10:30–11:00 Elias Willemse, Combining process mapping, business analytics

and operations research for effective problem solving (p. 38)

Parallel Session XIV: Vehicle Routing ProblemsChair: Angela Rademeyer [Venue: Jacaranda Room]

10:00–10:30 Kala Robert, The Residential Waste Collection Problem in an ur-ban area under constraints of restricted public infrastructure (p. 77)

10:30–11:00 Angela Rademeyer, Why travelling salesmen should considersleeping out (p. 95)

7

Page 12: Conference Theme: Better Decisions for Africa

Parallel Session XV: Decision MakingChair: Ozias Ncube [Venue: Orchid Room]

10:00–10:30 Ian Durbach, Behavioural decision making in sports predictions(p. 32)

10:30–11:00 Winnie Pelser, Decisions and analysis — What role does analysisplay? (p. 41)

Tea/Coffee (11:00–11:30) [Next to the Indoor Swimming Pool]

Tuesday 18 September 2012: (11:30–12:30)

Parallel Session XVI: Investment Portfolio AnalysisChair: Wim Gevers [Venue: Aloe Room]

11:30–12:00 Dave Evans, Meta modelling with linear programming for capitalinvestment with uncertain returns (p. 63)

12:00–12:30 Brian Mudhara, The management of investment portfolios usingoptimisation models (p. 62)

Parallel Session XVII: Problems in AgricultureChair: John Hearne [Venue: Jacaranda Room]

11:30–12:00 Sivashan Chetty, Results of local search heuristics for the AnnualCrop Planning Problem (p. 78)

12:00–12:30 Linke Potgieter, Modelling the impact and cost of the sterile in-sect technique on Eldana saccharina Walker in sugarcane (p. 65)

Parallel Session XVIII: Multi-criteria Decision AnalysisChair: Ian Durbach [Venue: Orchid Room]

11:30–12:00 Theodor Stewart, Developing spatial contiguity measures inmultiobjective land use planning (p. 44)

12:00–12:30 Muriel Chinoda, Using multi-criteria decision aid in corporateclimate change response (p. 93)

Lunch (12:30–13:45) [Main Dining Room]

Tuesday 18 September 2012: (13:45–15:15)

Parallel Session XIX: Optimisation in IndustryChair: Bernard Schlunz [Venue: Aloe Room]

13:45–14:15 Kizito Mubiru, Modelling and optimisation of production lotsizing decisions under stochastic demand: A case study involvinga milk powder product in Uganda (p. 67)

14:15–14:45 Kizito Mubiru, Modelling and optimisation of EOQ in super-markets under stochastic demand: A case study involving a milkpowder product in Uganda (p. 66)

14:45–15:15 David Zvipore, An operations management algorithm in the tilemanufacturing industry (p. 68)

8

Page 13: Conference Theme: Better Decisions for Africa

Parallel Session XX: Predictive ModellingChair: Khehla Moloi [Venue: Jacaranda Room]

13:45–14:15 Patricia Lutu, Replication of base models to improve the perfor-mance of positive vs negative classification ensembles for predictivedata mining (p. 76)

14:15–14:45 Gerbrand Breed, Practical application of semi-supervised seg-mentation within a predictive modelling context (p. 73)

14:45–15:15 Margarete Bester, Building predictive models using Xeno (p. 35)

Parallel Session XXI: Health CareChair: Tanya Lane-Visser [Venue: Orchid Room]

13:45–14:15 Nelishia Pillay & Christopher Rae, A survey of hyper-heuristics for the Nurse Rostering Problem (p. 89)

14:15–14:45 Nadia Viljoen, An implementable routing solution for home-basedcare in South Africa (p. 54)

14:45–15:15 Frans Snyders, Improving the work rate of community healthworkers through optimisation (p. 56)

Tea/Coffee (15:15–15:45) [Next to the Indoor Swimming Pool]

Tuesday 18 September 2012: (15:45–17:15)

ORSSA Annual General MeetingChair: Jan van Vuuren [Venue: Aloe Room]

Conference Banquet (19:00–) [Gorge Room]

Wednesday 19 September 2012

Wednesday 19 September 2012: (08:30–10:00)

Parallel Session XXII: Operations Research in DevelopmentChair: Marthi Harmse [Venue: Aloe Room]

08:30–09:00 Marthi Harmse, Community in operations research? (p. 39)09:00–09:30 Nadia Viljoen, Keeping it simple in a data-sparse environment:

The case of donor breastmilk demand and supply in South Africa(p. 60)

09:30–10:00 Hildah Mashira, An inventory routing system for dynamic fooddistribution (p. 59)

9

Page 14: Conference Theme: Better Decisions for Africa

Parallel Session XXIII: SchedulingChair: David Lubinsky [Venue: Jacaranda Room]

08:30–09:00 David Lubinsky, The science and magic of scheduling (p. 79)09:00–09:30 Robert Bennetto, Optimised schedules achieve massive savings

(p. 72)09:30–10:00 Colin Phillips, Call schedule design: Real world constraints and

technological solutions (p. 36)

Parallel Session XXIV: Statistical and Data AnalysisChair: Hannelie Nel [Venue: Orchid Room]

08:30–09:00 Caston Sigauke, Extreme daily increases in peak electricitydemand: Tail-quantile estimation (p. 49)

09:00–09:30 Hausitoe Nare, An econometric analysis of the effect of theZimbabwe Stock Exchange on the Zimbabwean economy in its yearsof decline (p. 46)

09:30–10:00 Farikayi Mutasa, A statistical analysis of monthly temperatureusing Box-Jenkins’ Arima methodology and a general linear modelapproach: A case study of the city of Bulawayo (p. 85)

Tea/Coffee (10:00–10:30) [Next to the Indoor Swimming Pool]

Wednesday 19 September 2012: (10:30–11:30)

Parallel Session XXV: Natural Resource ManagementChair: Mark Einhorn [Venue: Aloe Room]

10:30–11:00 Margarete Bester, Operational research in the forestry industry— Focussing on a recently developed allocation model (p. 69)

11:00–11:30 Paul Fatti, How healthy are the rhino populations in the Hluhluwe-iMfolosi Park? (p. 52)

Parallel Session XXVI: Information TechnologyChair: Hennie Kruger [Venue: Jacaranda Room]

10:30–11:00 Pieter Labuschagne, The feasibility of a generalised additiveneural network for spam classification (p. 50)

11:00–11:30 Erich Wilgenbus, A supervised learning approach for file fragmentclassification (p. 88)

Parallel Session XXVII: Transport and TransportationChair: Wessel Pienaar [Venue: Orchid Room]

10:30–11:00 Joke Buhrmann, Determining an optimal vehicle fleet mix (p. 43)11:00–11:30 Hannelie Nel, Using cluster analysis to extract trip and activity

information from GPS data (p. 91)

10

Page 15: Conference Theme: Better Decisions for Africa

Wednesday 19 September 2012: (11:40–13:00)

Plenary Session C: Conference ClosingChair: Theodor Stewart [Venue: Aloe Room]

11:40–12:40 John Hearne (Keynote Speaker)Conservation and the commercial utilisation of wildlife (p. 23)

12:40–12:55 Dave Evans, Paul Fatti & Hans Ittmann (Fellows of ORSSA)Reflection on Papers Read at the Conference

12:55–13:00 Jan van Vuuren (ORSSA President)Final Announcements, Thank Yous & Good bye

Lunch (13:00–14:00) [Main Dining Room]

11

Page 16: Conference Theme: Better Decisions for Africa

— List of Sessions & Chairs —

Ses

sion

Day

Tim

eS

lot

Top

icC

hai

rper

son

Ven

ue

AM

on

08:3

0–10

:00

Op

enin

gP

len

ary

Dav

eE

van

sA

loe

Room

BT

ue

08:3

0–0

9:5

0M

id-c

onfe

ren

ceP

len

ary

Pau

lF

atti

Alo

eR

oom

CW

ed11

:40–13

:00

Clo

sin

gP

len

ary

Th

eod

orSte

war

tA

loe

Room

IM

on10:3

0–12:3

0M

isce

llan

eou

sT

opic

sH

ans

Ittm

ann

Alo

eR

oom

IIM

on10

:30–1

2:3

0H

euri

stic

s&

Met

aheu

rist

ics

Eli

asW

ille

mse

Gor

geR

oom

III

Mon

10:3

0–1

2:3

0M

isce

llan

eou

sT

opic

sE

lias

Mu

nap

oJac

aran

da

Room

IVM

on10

:30–1

2:3

0T

acti

cal

Sit

uat

ion

Aw

aren

ess

Win

nie

Pel

ser

Orc

hid

Room

VM

on

13:4

5–15

:15

Sim

ula

tion

Lou

isD

ann

hau

ser

Alo

eR

oom

VI

Mon

13:4

5–1

5:1

5C

om

bin

ator

ial

Opti

mis

atio

nM

arga

rete

Bes

ter

Gor

geR

oom

VII

Mon

13:4

5–15

:15

Tec

hn

olog

yan

dE

du

cati

onD

anie

lL

otte

rJac

aran

da

Room

VII

IM

on

13:4

5–15

:15

Cas

eS

tud

yT

uto

rial

Nad

iaV

iljo

enO

rch

idR

oom

IXM

on

15:4

5–17

:15

Th

eP

etro

-ch

emic

alIn

du

stry

Pat

rick

Vel

dhu

izen

Alo

eR

oom

XM

on

15:

45–17:

15M

etah

euri

stic

s&

Art

ifici

alIn

tell

igen

ceA

der

emi

Ad

ewu

mi

Gor

geR

oom

XI

Mon

15:4

5–17

:15

Net

wor

ks

and

Con

nec

tivit

yT

jaar

tS

teyn

Jac

aran

da

Room

XII

Mon

15:

45–17:

15E

xec

uti

ve

Com

mit

tee

Mee

tin

gJan

van

Vu

ure

nO

rch

idR

oom

XII

IT

ue

10:

00–11:

00A

nal

yti

csC

asto

nS

igau

ke

Alo

eR

oom

XIV

Tu

e10:

00–11:

00V

ehic

leR

outi

ng

An

gela

Rad

emey

erJac

aran

da

Room

XV

Tu

e10:0

0–11:0

0D

ecis

ion

Mak

ing

Ozi

asN

cub

eO

rch

idR

oom

XV

IT

ue

11:

30–12:

30In

ves

tmen

tP

ortf

olio

An

alysi

sW

imG

ever

sA

loe

Room

XV

IIT

ue

11:3

0–12:3

0A

gric

ult

ure

Joh

nH

earn

eJac

aran

da

Room

XV

III

Tu

e11:3

0–12:3

0M

ult

i-cr

iter

iaD

ecis

ion

An

alysi

sIa

nD

urb

ach

Orc

hid

Room

XIX

Tu

e13:

45–15:

15O

pti

mis

atio

nin

Ind

ust

ryB

ern

ard

Sch

lun

zA

loe

Room

XX

Tu

e13:4

5–15:1

5P

red

icti

veM

od

elli

ng

Kh

ehla

Mol

oiJac

aran

da

Room

XX

IT

ue

13:

45–15:

15H

ealt

hC

are

Tan

yaL

ane-

Vis

ser

Orc

hid

Room

XX

IIW

ed08

:30–10

:00

Op

erati

ons

Res

earc

hin

Dev

elop

men

tM

arth

iH

arm

seA

loe

Room

XX

III

Wed

08:3

0–1

0:0

0S

ched

uli

ng

Dav

idL

ub

insk

yJac

aran

da

Room

XX

IVW

ed08:

30–10:

00S

tati

stic

alan

dD

ata

Anal

ysi

sH

ann

elie

Nel

Orc

hid

Room

XX

VW

ed10

:30–11

:30

Natu

ral

Res

ourc

eM

anag

emen

tM

ark

Ein

hor

nA

loe

Room

XX

VI

Wed

10:

30–11:

30In

form

atio

nT

ech

nol

ogy

Hen

nie

Kru

ger

Jac

aran

da

Room

XX

VII

Wed

10:

30–11:

30T

ran

spor

tan

dT

ran

spor

tati

onW

esse

lP

ien

aar

Orc

hid

Room

12

Page 17: Conference Theme: Better Decisions for Africa

— List of Papers —

(1) Airline taxi scheduling (Ian Campbell) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

(2) An algorithm for efficient secure network domination(Anton P de Villiers∗, Alewyn P Burger & Jan H van Vuuren) . . . . . . . . . . . 25

(3) Application of a harmony search algorithm to the core fuel reload optimisationproblem for the SAFARI-1 nuclear research reactor (Evert B Schlunz∗, PavelM Bokov & Rian H Prinsloo) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

(4) Application of optimisation methods to simulation models(Cecile Bezuidenhoudt) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

(5) Applying expert system technology in dietetics(Annette van der Merwe∗, Hennie Kruger & Tjaart Steyn) . . . . . . . . . . . . . . . 28

(6) An articulated qualitative model for evolutionary algorithms(Francois J Fagan∗ & Jan H van Vuuren) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

(7) An artificial bees colony algorithm for the Travelling Tournament Problem(Sandile Saul∗ & Aderemi O Adewumi) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

(8) An artificial neural network for detecting risk of Type 2 Diabetes(Benson Y Baha∗, Aderemi O Adewumi, Nachamanda V Blamah & GregoryM Wajiga) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .31

(9) Behavioural decision making in sports predictions (Ian Durbach) . . . . . . . . . 32

(10) The best posterior probability structure is not necessarily the causal structure(Jan W Kruger) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

(11) Blue Stallion Technologies Advertorial: How to save time and effort:Selected local case studies using the modelling and visualization tools of BlueStallion Technologies (Clemens Dempers) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

(12) Building predictive models using Xeno (Margarete J Bester) . . . . . . . . . . . . . . 35

(13) Call schedule design: Real world constraints and technological solutions(Colin A Phillips) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .36

(14) Coastal threat evaluation decision support(Jacques du Toit∗ & Jan H van Vuuren) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

(15) Combining process mapping, business analytics and operations research foreffective problem solving (Elias J Willemse) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

13

Page 18: Conference Theme: Better Decisions for Africa

(16) Community in operations research? (Martha FP Harmse) . . . . . . . . . . . . . . . . 39

(17) Closing Plenary: Conservation and the commercial utilisation ofwildlife (John W Hearne) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

(18) Consumer loan decisions with a profit-loss trade-off under multiple economicconditions (Kanshukan Rajaratnam∗ & Chun-Sung Huang) . . . . . . . . . . . . . . .40

(19) Decisions and analysis — What role does analysis play? (Winnie C Pelser) 41

(20) Design of a generic weapon assignment system in a ground-based air defenceenvironment (Daniel P Lotter∗ & Jan H van Vuuren) . . . . . . . . . . . . . . . . . . . . 42

(21) Determining an optimal vehicle fleet mix (Joke Buhrmann) . . . . . . . . . . . . . . .43

(22) Developing spatial contiguity measures in multiobjective land use planning(Theodor J Stewart) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

(23) Development and application of an assignment problem to reduce overcut wastein a secondary wood manufacturing facility (Berndt G Lindner∗ & Tanya Lane-Visser) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

(24) An econometric analysis of the effect of the Zimbabwe Stock Exchange onthe Zimbabwean economy in its years of decline (Hausitoe Nare∗ & AdrianT Ramhewa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

(25) The effect of network structure and switching costs on innovation diffusion insocial networks (Thandulwazi Magadla∗ & Ian Durbach) . . . . . . . . . . . . . . . . . 47

(26) Empirical results: Exact solutions using limited pattern generation for theN-sheet Cutting Stock Problem (Tjaart Steyn∗ & Giel Hattingh) . . . . . . . . . 48

(27) Extreme daily increases in peak electricity demand: Tail-quantile estimation(Caston Sigauke∗, Andrehette Verster & Delson Chikobvu) . . . . . . . . . . . . . . . 49

(28) The feasibility of a generalised additive neural network for spam classification(Pieter Labuschagne∗ & Tiny du Toit) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

(29) Finding the best pass-receiving position in the RoboCup Small-Size League(Moonyoung Yoon∗ & Tanya Lane-Visser) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .51

(30) How healthy are the rhino populations in the Hluhluwe-iMfolosi Park?(L Paul Fatti) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

(31) The hype around Analytics (Hans W Ittmann) . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

(32) An implementable routing solution for home-based care in South Africa(Nadia M Viljoen) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

14

Page 19: Conference Theme: Better Decisions for Africa

(33) Improved techniques to model continuous operations with discrete-eventsimulation (Anette van der Merwe) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

(34) Improving the work rate of community health workers through optimisation(Frans J Snyders∗ & Tanya Lane-Visser) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

(35) Integrating technology into teaching and learning in the school classroom(Olatunde O Osiyemi) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

(36) An intelligent particle swarm optimisation framework based on a multi-agentsystem (Nachamanda V Blamah∗, Aderemi O Adewumi, Gregory M Wajiga& Benson Y Baha) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

(37) An inventory routing system for dynamic food distribution(Hildah Mashira) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

(38) Keeping it simple in a data-sparse environment: The case of donor breastmilkdemand and supply in South Africa (Nadia M Viljoen∗, Julie Swann, MelihCelik, Wenwei Cao & Ozlem Ergun) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

(39) A linear response surface analysis approach to evaluate QoS in wirelessnetworks (Andre Brand∗, Hennie Kruger & Henry Foulds) . . . . . . . . . . . . . . . .61

(40) The management of investment portfolios using optimisation models(Brian Mudhara) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

(41) Meta modelling with linear programming for capital investment with uncertainreturns (Dave W Evans) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

(42) A model and results of local search heuristics for the Hostel Space AllocationProblem (Sunday A Ajibola & Aderemi O Adewumi∗) . . . . . . . . . . . . . . . . . . . .64

(43) Modelling the impact & cost of the sterile insect technique on Eldana saccha-rina Walker in sugarcane (Linke Potgieter∗ & Jan H van Vuuren) . . . . . . . . 65

(44) Modelling and optimisation of EOQ in supermarkets under stochastic demand:A case study involving a milk powder product in Uganda (Kizito P Mubiru∗,Kariko B Buhwezi & Peter Lating) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .66

(45) Modelling and optimisation of production lot sizing decisions under stochasticdemand: A case study involving a milk powder product in Uganda (Kizito PMubiru∗, Kariko B Buhwezi & Peter Lating) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

(46) An operations management algorithm in the tile manufacturing industry(David C Zvipore) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

(47) Operational research in the forestry industry — Focussing on a recentlydeveloped allocation model (Margarete J Bester) . . . . . . . . . . . . . . . . . . . . . . . . . . 69

15

Page 20: Conference Theme: Better Decisions for Africa

(48) An optimal distribution strategy for a supply chain organisation: The case ofDelta Beverages (Absalom Jaison∗, Desmond Mwembe∗ & Justin Chirima) 70

(49) The optimisation of terrain modelling and line-of-sight techniques(Andries M Heyns∗ & Jan H van Vuuren) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

(50) Optimised schedules achieve massive savings (Robert A Bennetto) . . . . . . . .72

(51) Practical application of semi-supervised segmentation within a predictivemodelling context (Gerbrand D Breed∗, Tanja de la Rey & Stephanus ETerblanche) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .73

(52) A practical application of stochastic simulation within an explosives supplychain environment (Jaco L Joubert∗ & Gerhard Carstens) . . . . . . . . . . . . . . . .74

(53) A problem reduction approach for the Survivable Network Design Problem(Stephanus E Terblanche) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

(54) Replication of base models to improve the performance of positive vs negativeclassification ensembles for predictive data mining (Patricia EN Lutu) . . . . 76

(55) The Residential Waste Collection Problem in an urban area underconstraints of restricted public infrastructure (Kala KJ Robert∗ & MourdounT Christelle) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .77

(56) Results of local search heuristics for the Annual Crop Planning Problem(Sivashan Chetty∗ & Aderemi O Adewumi) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

(57) Mid-conference Plenary: The Sasol group energy linear program-ming model (Diki IJ Langley & Patrick R Veldhuizen∗) . . . . . . . . . . 22

(58) The science and magic of scheduling (David J Lubinsky) . . . . . . . . . . . . . . . . . 79

(59) Self-organising traffic light control(Mark D Einhorn∗, Jan H van Vuuren & Alewyn P Burger) . . . . . . . . . . . . . . 80

(60) Solving a binary linear programming model in polynomial time(Elias Munapo) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

(61) Opening Plenary: Spatial problems in managing land for habitatand fire (John W Hearne) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

(62) Spreadsheet solvers: A comparison (Wim R Gevers) . . . . . . . . . . . . . . . . . . . . . . 82

(63) Stochastic simulation for Sasol Solvents Global Tank Planning(Ester J Vermaak∗, Leilani E Meijer & Louis H Snyders) . . . . . . . . . . . . . . . . . 83

(64) Strategic coordination in a production supply chain (Tichoana C Mazuru) 84

16

Page 21: Conference Theme: Better Decisions for Africa

(65) A statistical analysis of monthly temperature using Box-Jenkins’ Arima method-ology and a general linear model approach: A case study of the city ofBulawayo (Farikayi K Mutasa∗, Nonhlanhla Magadlela, Philimon Nyamugure& Edward T Chiyaka) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

(66) Studies in metaheuristics for the Blood Assignment Problem(Emmanuel Dufourq, Michael O Olusanya∗ & Aderemi O Adewumi) . . . . . 86

(67) A successful case study for integration between business and analytics in thePropylene value chain (Nica van der Westhuizen) . . . . . . . . . . . . . . . . . . . . . . . . 87

(68) A supervised learning approach for file fragment classification(Erich Wilgenbus∗, Hennie Kruger & Tiny du Toit) . . . . . . . . . . . . . . . . . . . . . . 88

(69) A survey of hyper-heuristics for the Nurse Rostering Problem(Nelishia Pillay∗ & Christopher Rae∗) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

(70) Teachers’ and learners’ perception on mathematics literacy as subject stream(Olatunde O Osiyemi) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

(71) Using cluster analysis to extract trip and activity information from GPS data(Johanna H Nel∗ & Stephan C Krygsman) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

(72) Case Study Tutorial: Using operations research for strategic planning in thenot-for-profit health sector — A case study (Nadia M Viljoen∗, Wenwei Cao,Melih Celik, Julie Swann & Ozlem Ergun) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

(73) Using multi-criteria decision aid in corporate climate change response(Muriel Chinoda∗ & Jan W Kruger) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

(74) The weapon assignment scheduling problem in a ground-based air defenceenvironment (Michelle van der Merwe∗ & Jan H van Vuuren) . . . . . . . . . . . . 94

(75) Why travelling salesmen should consider sleeping out(Angela L Rademeyer∗, Robert Bennetto & Alexander Sloan) . . . . . . . . . . . . 95

17

Page 22: Conference Theme: Better Decisions for Africa

— List of Authors —

Adewumi, Aderemi O (University of KwaZulu-Natal, South Africa) 30, 31, 58, 64, 78, 86Ajibola, Sunday A (University of KwaZulu-Natal, South Africa) . . . . . . . . . . . . . . . . . . . . 64Baha, Benson Y (Mainstreet Bank, Nigeria) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31, 58Bennetto, Robert A (OPSI Systems, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72, 95Bester, Margarete J (XTranda, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .35, 69Bezuidenhoudt, Cecile (University of Cape Town, South Africa) . . . . . . . . . . . . . . . . . . . 27Blamah, Nachamanda V (University of KwaZulu-Natal, South Africa) . . . . . . . . . . . 31, 58Bokov, Pavel M (South African Nuclear Energy Corporation) . . . . . . . . . . . . . . . . . . . . . . . 26Brand, Andre (North-West University, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .61Breed, Gerbrand D (North-West University, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . 73Buhrmann, Joke (OPSI Systems, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43Buhwezi, Kariko B (Makerere University, Uganda) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66, 67Burger, Alewyn P (Stellenbosch University, South Africa) . . . . . . . . . . . . . . . . . . . . . . . 25, 80Campbell, Ian (University of Witwatersrand, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . 24Carstens, Gerhard (Sasol Shared Services, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . .74Cao, Wenwei (Georgia Institute of Technology, United States of America) . . . . . . . . 60, 92Celik, Melih (Georgia Institute of Technology, United States of America) . . . . . . . . 60, 92Chetty, Sivashan (University of KwaZulu-Natal, South Africa) . . . . . . . . . . . . . . . . . . . . . . 78Chikobvu, Delson (University of the Free State, South Africa) . . . . . . . . . . . . . . . . . . . . . . 49Chinoda, Muriel (University of South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93Chirima, Justin (Great Zimbabwe University, Zimbabwe) . . . . . . . . . . . . . . . . . . . . . . . . . . . 70Chiyaka, Edward T (NUST, Zimbabwe) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85Christelle, Mourdoun T (Catholic University of Central Africa, Cameroon) . . . . . . . . . 77De la Rey, Tanja (North-West University, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . .73De Villiers, Anton P (Stellenbosch University, South Africa) . . . . . . . . . . . . . . . . . . . . . . . 25Dempers, Clemens (Blue Stallion Technologies, South Africa) . . . . . . . . . . . . . . . . . . . . . . 34Du Toit, Jacques (Stellenbosch University, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . .37Du Toit, Tiny (North-West University, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50, 88Dufourq, Emmanuel (University of KwaZulu-Natal, South Africa) . . . . . . . . . . . . . . . . . . 86Durbach, Ian (University of Cape Town, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . 32, 47Einhorn, Mark D (Stellenbosch University, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . 80Ergun, Ozlem (Georgia Institute of Technology, United States of America) . . . . . . .60, 92Evans, Dave W (Development Bank of Southern Africa, South Africa) . . . . . . . . . . . . . . 63Fagan, Francois J (Stellenbosch University, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . 29Fatti, L Paul (University of the Witwatersrand, South Africa) . . . . . . . . . . . . . . . . . . . . . . . 52Foulds, Henry (North-West University, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61Gevers, Wim R (Stellenbosch University, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82Harmse, Martha FP (Sasol Synfuels, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39Hattingh, Giel (North-West University, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48Hearne, John W (Royal Melbourne Institute of Technology, Australia) . . . . . . . . . . 21, 23Heyns, Andries M (Stellenbosch University, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . .71

18

Page 23: Conference Theme: Better Decisions for Africa

Huang, Chun-Sung (University of Cape Town, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . 40Ittmann, Hans W (HWI Consulting, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53Jaison, Absalom (NUST, Zimbabwe) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .70Joubert, Jaco L (Sasol Shared Services, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74Kruger, Jan W (University of South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33, 93Kruger, Hennie (North-West University, South Africa) . . . . . . . . . . . . . . . . . . . . . . . 28, 61, 88Krygsman, Stephan C (Stellenbosch University, South Africa) . . . . . . . . . . . . . . . . . . . . . . 91Labuschagne, Pieter (North-West University, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . 50Lane-Visser, Tanya E (Stellenbosch University, South Africa) . . . . . . . . . . . . . . . . 45, 51, 56Langley, Diki IJ (Sasol Shared Services, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22Lating, Peter (Makerere University, Uganda) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66, 67Lindner Berndt G (Stellenbosch University, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . 45Lotter Daniel P (Stellenbosch University, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42Lubinsky David J (OPSI Systems, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79Lutu Patricia EN (University of Pretoria, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76Magadla, Thandulwazi (University of Cape Town, South Africa) . . . . . . . . . . . . . . . . . . . . 47Magadlela, Nonhlanhla (NUST, Zimbabwe) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85Mashira, Hildah (NUST, Zimbabwe) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59Mazuru, Tichoana C (NUST, Zimbabwe) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84Meijer, Leilani E (Sasol Technology, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83Mubiru, Kizito P (Kyambogo University, Uganda) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66, 67Mudhara, Brian (NUST, Zimbabwe) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62Munapo, Elias (University of South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .81Mutasa, Farikayi K (NUST, Zimbabwe) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85Mwembe, Justin (NUST, Zimbabwe) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70Nare, Hausitoe (NUST, Zimbabwe) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46Nel, Johanna H (Stellenbosch University, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91Nyamugure, Philimon (NUST, Zimbabwe) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85Olusanya, Michael O (University of KwaZulu-Natal, South Africa) . . . . . . . . . . . . . . . . . . 86Osiyemi, Olatunde O (University of Fort Hare, South Africa) . . . . . . . . . . . . . . . . . . . . 57, 90Pelser, Winnie C (Armscor, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41Phillips, Colin A (OPSI Systems, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .36Pillay, Nelishia (University of KwaZulu-Natal, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . 89Potgieter, Linke (Stellenbosch University, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65Prinsloo, Rian H (South African Nuclear Energy Corporation) . . . . . . . . . . . . . . . . . . . . . .26Rademeyer, Angela L (OPSI Systems, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95Rae, Christopher (University of KwaZulu-Natal, South Africa) . . . . . . . . . . . . . . . . . . . . . . 89Rajaratnam, Kanshukan (University of Cape Town, South Africa) . . . . . . . . . . . . . . . . . .40Ramhewa, Adrian T (Private Capacity, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46Robert, Kala KJ (Catholic University of Central Africa, Cameroon) . . . . . . . . . . . . . . . . 77Saul, Sandile (University of KwaZulu-Natal, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . 30Schlunz, Evert B (South African Nuclear Energy Corporation) . . . . . . . . . . . . . . . . . . . . . .26Sigauke, Caston (University of Limpopo, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49Sloan, Alexander (OPSI Systems, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .95Snyders, Frans J (Stellenbosch University, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56Snyders, Louis H (Sasol Technology, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .83Stewart Theodor J (University of Cape Town, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . 44

19

Page 24: Conference Theme: Better Decisions for Africa

Steyn, Tjaart (North-West University, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .28, 48Swann, Julie (Georgia Institute of Technology, United States of America) . . . . . . . . 60, 92Terblanche, Stephanus E (North-West University, South Africa) . . . . . . . . . . . . . . . . .73, 75Van der Merwe, Anette (Sasol Technology, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . 55Van der Merwe, Annette (North-West University, South Africa) . . . . . . . . . . . . . . . . . . . 28Van der Merwe, Michelle (Stellenbosch University, South Africa) . . . . . . . . . . . . . . . . . . 94Van der Westhuizen, Nica (Sasol Shared Services, South Africa) . . . . . . . . . . . . . . . . . . 87Van Vuuren, Jan H (Stellenbosch University, South Africa) . 25, 29, 37, 42, 65, 71, 80, 94Veldhuizen, Patrick R (Sasol Shared Services, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . 22Vermaak, Ester J (Sasol Technology, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83Verster, Andrehette (University of the Free State, South Africa) . . . . . . . . . . . . . . . . . . . . 49Viljoen, Nadia M (Council for Scientific & Industrial Research, South Africa) 54, 60, 92Wajiga, Gregory M (Modibbo Adama University of Technology, Nigeria) . . . . . . . . 31, 58Wilgenbus, Erich (North-West University, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . .88Willemse, Elias J (LTS Consulting, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38Yoon, Moonyoung (Stellenbosch University, South Africa) . . . . . . . . . . . . . . . . . . . . . . . . . . . 51Zvipore, David C (NUST, Zimbabwe) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

20

Page 25: Conference Theme: Better Decisions for Africa

— Plenary Paper Abstracts —

Opening Plenary:

Spatial problems in managing land for habitat and fire

John W HearneSchool of Mathematical and Geospatial SciencesRoyal Melbourne Institute of Technology, [email protected]

Abstract

The selection of land for various purposes such as recreation, development or con-servation can usually be accomplished with a simple model that maximises someattribute subject to a cost constraint. Alternatively, cost is minimised subject toa constraint that ensures sufficient quality of an attribute. This problem is moredifficult to solve when selected parcels of land have to be connected as in the ‘reservedesign’ problem. There are various heuristic methods for solving this problem, butthey do not guarantee an optimal solution. Integer programming methods gener-ally involve ‘sub-tours’ that have to be eliminated in a non-automated and iterativemanner. A new automated solution approach to this problem which generates aglobal optimum will be presented and illustrated.

Wildlife that roam freely amongst connected properties that are individually ownedpresent problems. These relate to the equitable distribution of costs and benefitsarising from their presence. Two case studies will be discussed: Kangaroo and do-mestic stock in outback Queensland; and neighbours following opposing consumptiveand non-consumptive uses of African wildlife.

Finally, the devastating fires that killed more than a hundred people in Victoria in2010 will be discussed. Recommendations made by the Royal Commission investi-gating the event are best implemented using operations research. Where should fuelload be reduced to minimise the risk of loss given limited resources? Where shouldlimited fire-fighting resources be located? What acquisitions (e.g. trucks, aircraft)should be made given limited budgets? Solutions to some of these questions will beproposed and illustrated.

21

Page 26: Conference Theme: Better Decisions for Africa

Mid-conference Plenary:

The Sasol group energy linear programming model

Diki IJ Langley & Patrick R Veldhuizen∗

Sasol Shared Services, A Division of [email protected] & [email protected]

Abstract

The Sasol group energy linear programming model was showcased in the INFORMSPrize nomination package and contributed to Sasol winning the prize in 2011. Itwas developed in response to rising electricity costs. Previously the largest eco-nomic drivers in company linear programmes were raw material costs and productsales prices. Electricity costs had to be included and this resulted in some widerconsiderations and scoping of the work:

• Electricity is not a ‘stand-alone’ utility, as it is not only bought in from anexternal provider, but is also produced on some of the production sites bysteam-driven turbines. It is thus part of an integrated site energy balancewhich, in turn, is integrated with facility material flows and balances.

• Until that time the group linear program only contained two of the six pro-duction facilities. The others had to be included in order to calculate impactof electricity pricing and supply on Sasol South Africa as a whole. This thenresulted in a group linear program which, because it now had all the sitesincluded, could be used as a strategic planning tool.

The resulting group energy linear programming model is the only tool within thegroup that combines and inter-relates chemical/molecular and energy flows on agroup-wide scale and currently is used on a regular basis for group sales and oper-ability planning which set directives for business units. It is also increasingly beingused to evaluate:

• Proposed new units or unit modifications,

• Future company material and energy flow balances,

• Sensitivities of rising costs and prices, and

• Economic debottlenecking.

22

Page 27: Conference Theme: Better Decisions for Africa

Closing Plenary:

Conservation and the commercial utilisation of wildlife

John W HearneSchool of Mathematical and Geospatial SciencesRoyal Melbourne Institute of Technology, [email protected]

Abstract

A couple of decades ago a translocation strategy for saving the endangered blackrhino was formulated. A model was developed to assist with decision-making aroundthe rate of translocation for each age-class. Some of the subsequent implementationproblems of the strategy will be discussed. It will be shown how some issues havebeen addressed by including the perspectives of two different groups of managers in anew model. It has been argued that the commercial utilisation of wildlife offers theirbest prospects for survival. In South Africa this industry has grown substantiallyover the last two decades. Various optimisation models have been used to determinethe mix of species and stocking rates that will maximise returns from game onthese private reserves. These include simple LPs to portfolio optimisation, and‘packaging.’ Some examples will be discussed and future modelling needs suggested.

23

Page 28: Conference Theme: Better Decisions for Africa

— Contributed Paper Abstracts —

Airline taxi scheduling

Ian CampbellUniversity of the Witwatersrand, South [email protected]

Abstract

An airline taxi scheduling problem is described as experienced by a tourist airlineoperating in the Okavango Delta, Botswana. Typically a daily schedule is drawn upmanually by a pair of experienced schedulers a few days before the day in question.In this research, the problem is modelled and optimised using a multi-commoditynetwork flow model implemented in Matlab and CPLEX. A series of heuristics aredescribed to reduce the problem size such that it can be solved within a reasonabletime on a personal computer. A large daily schedule of booking requests was suppliedby the airline. The model produced a saving of 11% of aircraft operating costs whencompared with the manual schedule that was actually used on that particular day.

24

Page 29: Conference Theme: Better Decisions for Africa

An algorithm for efficient secure network domination

Anton P de Villiers∗, Alewyn P Burger & Jan H van VuurenStellenbosch University, South [email protected], [email protected] & [email protected]

Abstract

A dominating set of a network is a subset of nodes in the network at which guardsmay be placed such that each node of the network is either in the subset (i.e.occupied by a guard) or adjacent to a node in the subset (i.e. has a neighbouringguard). The domination number of a network is the size of a smallest dominatingset of the network. A secure dominating set of a network, on the other hand, is adominating set with the additional property that it should be possible for a guardto leave his current node u and move to an adjacent node v (e.g. when a securityconcern arises at the location modelled by v), with the resulting guard placement(after the move) also being a dominating set of the network. The secure dominationnumber of a network is the smallest number of guards in a secure dominating set ofthe network.

A recursive algorithm is developed for computing the secure domination number of anetwork. The algorithm relies on general bounds on the secure dominating numberof the network, as well as a characterisation of secure dominating sets in networks.The time complexity of this algorithm is examined and results obtained by meansof the algorithm for a number of classes of networks are also presented.

25

Page 30: Conference Theme: Better Decisions for Africa

Application of a harmony search algorithm to the core fuelreload optimisation problem for the SAFARI-1 nuclearresearch reactor

Evert B Schlunz∗, Pavel M Bokov & Rian H PrinslooThe South African Nuclear Energy Corporation, South [email protected], [email protected] &[email protected]

Abstract

Nuclear reactors are typically operated at power for a given duration, which isfollowed by a shut down period. One of the tasks that occurs during the shutdown period is the reloading of fuel assemblies in the core in preparation for thenext cycle. Typically, a number of depleted fuel assemblies are removed from thecore and replaced by fresh fuel assemblies. The loading positions of all the fuelassemblies in the core may then be reconfigured in order to satisfy prescribed safetyand utilisation requirements.

The core fuel reload optimisation problem (CFROP) refers to the problem of findingan optimal fuel loading configuration for a nuclear reactor core. The CFROP is typ-ically multiobjective, nonlinear, discrete and combinatorial in nature. Furthermore,a reactor core calculational system is used to evaluate the objective function(s) ofthe CFROP for a given fuel loading configuration and entails the calculation of thetemporal, spatial and spectral distribution of neutron flux in the reactor model.

The South African Nuclear Energy Corporation (Necsa) operates South Africa’s nu-clear research reactor, known as the South African Fundamental Atomic ResearchInstallation 1 (SAFARI-1). SAFARI-1 is primarily utilised for scientific research,radiopharmaceutical and isotope production, irradiation services and material test-ing.

In this paper, a mathematical formulation of the CFROP will be given, incorpo-rating the objectives and constraints applicable to the operation of the SAFARI-1reactor. Issues arising from the complexity and multiobjective nature of the CFROPfor SAFARI-1 will be discussed, followed by a description of the harmony search al-gorithm that has been implemented to solve the problem approximately. A keyaspect considered is the limited computational budget within which a solution mustbe obtained. The results that were obtained and a brief discussion on future researchwill conclude the paper. The aim of the study is to develop an efficient approach topropose good fuel loading configurations from cycle to cycle.

26

Page 31: Conference Theme: Better Decisions for Africa

Application of optimisation methods to simulation models

Cecile BezuidenhoudtUniversity of Cape Town, South [email protected]

Abstract

This paper investigates the potential for the application of heuristic optimisationmethods to goal seeking in simulation models. The aim is to provide a tool toinform purchasing policy design. Three heuristic methods (genetic algorithms, sim-ulated annealing and nested partitions) were applied to a simple hypothetical testcase simulation model based broadly on the functioning and gas consumption of anOpen Cycle Gas Turbine plant for electricity generation. The investigation consid-ers policy decisions on fuel orders based on forecasted demand for some months intothe future, after which actual demands are simulated to generate changes in stocks,fuel shortages and unnecessary burning of excess fuels. The policies considered bythe model relate decisions made on fuel orders to parameters of the probability dis-tribution on forecasted demand. Each optimisation technique is considered in turnto see how the inherent randomness in the simulation model is handled and thenthe methods are then compared. Potential for application to a more comprehensiveelectricity generation system will also be discussed

27

Page 32: Conference Theme: Better Decisions for Africa

Applying expert system technology in dietetics

Annette van der Merwe∗, Hennie Kruger & Tjaart SteynNorth-West University, South [email protected], [email protected]& [email protected]

Abstract

An expert system is an intelligent computer program which uses inference proceduresand knowledge to solve problems that require advanced human expertise. Expertsystems are used in a vast variety of applications, and are readily available on theinternet, providing individualised solutions to specific problems. However, in certainfields where several unknown variables contribute to an optimal solution, few suchsystems are available.

A study is being conducted to determine the feasibility of combining expert systemtechnology with linear programming techniques to enable solving these types ofproblems. The application area of dietetics provides an ideal experimental base forsuch a study. Although many diet systems are already freely available, they solveonly a partial section of the problem as a whole. In this application area, an expertsystem is proposed which uses heuristics and an advanced knowledge base to yielda linear program model. The model is then used to solve the problem of providinga balanced eating plan for an individual based on responses made to the expertsystem.

The literature indicates that unbalanced eating habits are the cause for many health-related issues in South Africa. It is believed that a system like the one proposedhere might aid in establishing better health by allowing people access to expertknowledge.

Progress on the knowledge engineering process and development of the expert systemwill be presented.

28

Page 33: Conference Theme: Better Decisions for Africa

An articulated qualitative model for evolutionary algorithms

Francois J Fagan∗ & Jan H van VuurenStellenbosch University, South [email protected] & [email protected]

Abstract

The fitness landscape has long been the dominant paradigm of metaheuristic opti-misation algorithms. Along with intuitive terms such as exploitation, exploration,diversity and intensity, qualitative descriptions of models are used to complementquantitative results. These descriptions largely fail due to imprecise (often con-tradictory) terminology and the lack of a formal articulated qualitative model. Aformal articulated qualitative model for evolutionary algorithms on a continuoussearch space is presented, which extends the notion of a fitness landscape to a popu-lation fitness landscape. This is used to precisely define relevant terminologies, suchthat they agree with prevalent opinions in the literature , and construct the idealprobability distribution for analysing evolutionary operators.

29

Page 34: Conference Theme: Better Decisions for Africa

An artificial bees colony algorithm for the TravellingTournament Problem

Sandile Saul∗ & Aderemi O AdewumiUniversity of KwaZulu-Natal, South [email protected] & [email protected]

Abstract

Scheduling of professional sports is one of many researched practical problems incombinatorial optimisation. The scheduling of professional sports is a known NP-Hard problem which is very difficult to solve as it involves multiple constraints.This paper addresses the Travelling Tournament Problem (TTP). The goal of TTPis to create a feasible sport schedule that minimises the distance traveled by teams.An Artificial Bee Colony (ABC) algorithm was designed for the problem and thealgorithm was applied to different instances of the problem. Results obtained arecompared with previous results from the literature.

30

Page 35: Conference Theme: Better Decisions for Africa

An artificial neural network for detecting risk ofType 2 Diabetes

Benson Y Baha∗

Information Technology & Systems, Mainstreet Bank, [email protected]

Aderemi O Adewumi & Nachamanda V BlamahUniversity of KwaZulu-Natal, South [email protected] & [email protected]

Gregory M WajigaModibbo Adama University of Technology, [email protected]

Abstract

Diabetes is a chronic lifelong disease characterised by hyper-glycaemia resultingfrom defects in insulin secretion, insulin action, or both, which increases the risksof long-term damage, dysfunction and failure of various organs — especially theeyes, kidneys, nerves, heart, gallbladder or blood vessels. Type 2 Diabetes (T2D)constitutes 85–90% of all cases of diabetes with an expectation of 300 million casesaround the world by the year 2025. These complications and high prevalence of T2Dnecessitated the study. This paper proposes a new model for identifying individualsat high risk of developing T2D. Data were collected from three different sources.The first is from epidemiological studies with the help of a diabetologist in which13 risk factors associated with development of T2D were identified. Professionalrespondents using structured liket format to ascertain the degree of severity of riskfactors. The last source of data was from non-professional respondents for training,validation and testing the network. The result obtained from AHP technique wasused to proffer a new artificial neural network model that could identify individualsat high risk of developing T2D. The best performed network identified during thetraining contained two hidden layers consisting of six and two neurons, respectively,and an output layer consisting of one neuron. The network recorded best validationperformance of 0.1000 at 532th epoch and a correlation coefficient of 0.9981. Aregression plot indicated exact linear relationships with all the axes close to 1. Atleast 266 out of 1 122 cases in the data set used were found to be close to 1, whichindicated high risk of T2D.

31

Page 36: Conference Theme: Better Decisions for Africa

Behavioural decision making in sports predictions

Ian DurbachUniversity of Cape Town, South [email protected]

Abstract

Behavioural decision making attempts to study patterns and processes in humanjudgement and choice. Many findings have been used to improve the processes bywhich judgmental inputs are obtained from decision makers — in decision analysisand decision support systems, for example. This paper presents the results of ananalysis of the forecasting behaviour of some 80 000 users of a sports prediction web-site. In particular we examine how surprising outcomes affect future expectations,and whether (and under what conditions) a “wisdom of the crowds” phenomenoncan arise.

32

Page 37: Conference Theme: Better Decisions for Africa

The best posterior probability structure is not necessarilythe causal structure

Jan W KrugerUniversity of South Africa, South [email protected]

Abstract

A Bayesian belief network structure with the highest posterior probability is notnecessarily the causal structure. The graphical language used in machine learning isdescribed. The Cooper and Herskovits equation is given, followed by proof that themaximum posterior probability structure is not the same as the causal structure. Ina specific example the Cooper and Herskovits algorithm is tested to see how oftenthe causal structure is found.

33

Page 38: Conference Theme: Better Decisions for Africa

Blue Stallion Technologies Advertorial:How to save time and effort: Selected local case studiesusing the modelling and visualization tools of Blue StallionTechnologies

Clemens DempersBlue Stallion Technologies, South [email protected]

Abstract

Using an appropriate tool can save operations research practitioners a lot of time andeffort. Selected examples from local consulting projects and applications will be pre-sented to illustrate the analytical tools distributed by Blue Stallion Technologies inSouth Africa. This will include simulation modelling, optimisation, management re-porting and decision support. And the products demonstrated will include Anylogic,Mathematica, Lingo and Expert Choice.

34

Page 39: Conference Theme: Better Decisions for Africa

Building predictive models using Xeno

Margarete J BesterXTranda, South [email protected]

Abstract

The purpose of this paper is to introduce the scorecard building capabilities ofXeno. There are a number of different scorecard development tools in the market.Traditionally scorecards were built using various regression techniques, but recentlysoftware came to the market using mathematical programming or goal programmingtechniques. It is unusual to find a single performance definition that describes thetotal outcome of a business decision. Traditionally multiple scorecards were usedto address this problem, but Xeno has introduced a new technique using multipleperformance definitions building one scorecard. In this talk the various techniquesused in Xeno will be presented in a demo and the new feature of using multipleperformances in building a single model will be discussed.

35

Page 40: Conference Theme: Better Decisions for Africa

Call schedule design: Real world constraints andtechnological solutions

Colin A PhillipsOPSI Systems, South [email protected]

Abstract

Call schedules allocate reps to visit sites in order to minimise distance and balanceworkload over long repeated cycles. Recent innovations in software allow for suchcall schedules to be created quickly and updated quickly, and also ensure that realworld constraints are maintained.

A major producer of consumer goods had over 11 000 sites to be called upon in oneprovince of South Africa, where each region was managed separately. A restructuringof the business required that all call schedules be updated to handle an increasein reps, an increase in calling frequency, changes in calling times, and additionalbusiness constraints.

This report focuses on the methodology and software solution used to provide asatisfactory solution in a short amount of time.

36

Page 41: Conference Theme: Better Decisions for Africa

Coastal threat evaluation decision support

Jacques du Toit∗ & Jan H van VuurenStellenbosch University, South [email protected] & [email protected]

Abstract

Advances in technology are steadily contributing to improved maritime domainawareness. South Africa actively monitors AIS (Automatic Identification System)and VMS (Vessel Monitoring System) transmissions and gathers additional datathrough a number of strategically placed coastal radars. Surveillance systems pro-vide maritime domain operators and analysts with vast quantities of data from whichthe operators must distil the relevant situational facts of interest. Human opera-tors perform these tasks through reliance on experience and domain knowledge assituational facts must typically be inferred from the data.

Support tools play a valuable role in this context by processing these vast collectionsof data, identifying concepts of interest and predicting future occurrences. It isessential that such a tool is flexible enough to adapt to changes in vessel behaviourand that it is capable of integrating new behaviours. A framework for such a systemis presented in this talk where the emphasis is on assisting an operator within asurveillance context. A few system components will be discussed in greater detailand experimental results will be presented.

37

Page 42: Conference Theme: Better Decisions for Africa

Combining process mapping, business analytics andoperations research for effective problem solving

Elias J WillemseLTS Consulting, South [email protected]

Abstract

Industrial engineers are equipped with a wide range of problem solving tools, rang-ing from analytical operations research (OR) tools that can be applied to specificwell-defined problems, to more general high level tools, such as business processmapping that can be applied across a whole organisation. The advantages andshortcomings of both extremities are well documented. OR is cited for being tootechnical and narrow-minded, while business process mapping is criticised for yield-ing only marginal improvements. Recently business analytics has emerged as acomplementary approach to traditional OR methods. The approach aims to betterunderstand business performance by using data mining in combination with quan-titative analysis and is regarded more robust than OR methods. Nonetheless it toois criticised as transactional data alone cannot fully address business performance.

This presentation demonstrates, through a case study, how the strengths of thedifferent approaches can be leveraged for effective problem solving. The case studyfocuses on improving operations at a wholesale plant nursery in the Western Cape,South Africa. It illustrates how the three approaches were integrated, and discussesthe benefits of combining the approaches.

38

Page 43: Conference Theme: Better Decisions for Africa

Community in operations research?

Martha FP HarmseSasol Synfuels, South [email protected]

Abstract

Most attempts to give a balanced and representative account of what operationsresearch entails, are rather lengthy [1]. It is similarly challenging to delimit whatconstitutes community-based operations research [2]. A contributing factor is thediverse perspectives about what establishes a community. This paper explores oneview about a community and the implications for community-based operations re-search, but focuses on suggestions for operations research in general. The paper isstructured around a specific case study.

[1] Operations Research Society Of South Africa, What is operations re-search?, [Online], [Cited 29 July 2012], Available from http://www.orssa.org.za/

wiki/pmwiki.php?n=Main.WhatIsOR

[2] Bryant J, Ritchie C & Taket A, 1994, Messages for the OR practitioner,pp. 231–240 in Ritchie C, Taket A & Bryant J (Eds.), Community works,PAVIC, Sheffield.

39

Page 44: Conference Theme: Better Decisions for Africa

Consumer loan decisions with profit-loss trade-off undermultiple economic conditions

Kanshukan Rajaratnam∗ & Chun-Sung HuangUniversity of Cape Town, South [email protected] & [email protected]

Abstract

A topic of recent interest is loss prevention and cash-flow management in consumerloan portfolios. Past literature considers the case of a portfolio manager with multi-ple but conflicting objectives, such as maximising profits, maximising market share,and minimising risk. This was later extended to incorporate notions of economicconditions where a portfolio manager is faced with making a decision prior to the re-alisation of future economic conditions. For example, the portfolio manager is facedwith trade-off between maximising expected profit and market share, under the pos-sibility of one out of two future economic conditions. However, the accept/rejectdecision must be made prior to the realisation of an economic condition. In thispaper, we extend the notion of multiple economic conditions to the expected profitand expected loss space. We limit our discussion to a portfolio manager with accessto a single scorecard, but with customer performance differing under two differenteconomic conditions. We show that all operating points on the efficient frontier is aresult of single cutoff-score policy.

40

Page 45: Conference Theme: Better Decisions for Africa

Decisions and Analysis — What role does analysis play?

Winnie C PelserArmscor, South [email protected]

Abstract

In the strategic defence environment decisions are made constantly. For us as ana-lysts the important question is: Does analysis help? One of the important reasonsfor employing analysis is to achieve evidence-based decision making.

Strategic decision making is difficult because the future is always uncertain, theenvironment is complex and there is always interdependence between factors andpossible decisions. Since high-level decisions are increasingly subjective and increas-ingly uncertain, analysis is also vital because it will determine the cause of futurepolicies or posture. There is always a playoff between the rate of producing analysisand the rate of progress in context, decision or question. This leads to a push vspull situation. The push refers to analysis done that is not confined to a pre-definedquestion. The pull refers to pre-defined questions.

The important question remains: Does analysis help to make sound and timelydecisions?

41

Page 46: Conference Theme: Better Decisions for Africa

Design of a generic weapon assignment system in a ground-based air defence environment

Daniel P Lotter∗ & Jan H van VuurenStellenbosch University, South [email protected] & [email protected]

Abstract

In a typical ground-based air defence system (GBADS), defended assets (DAs) requireprotection from a number of ground-based weapon systems (WSs) against aerialvehicles entering the defended airspace. These aerial vehicles have to be classi-fied as threats and the levels of threat they pose to DAs have to be established.This information is then used to assign one or a number of available WSs to en-gage the threats according to a pre-specified criterion, a problem which is known asthe weapon assignment (WA) problem. A threat evaluation and weapon assignment(TEWA) decision support system may be employed to aid military personnel inthe complex process of classification and evaluation of threats and the subsequentassignment of WSs to threats.

The focus in this paper is on the WA subsystem of the larger TEWA system describedabove, where the aim is to put forward a first-order generic WA subsystem archi-tecture. This framework includes a specification of the data required by a WAsubsystem, the components contained within a WA subsystem, the processes in-volved in the working of such a subsystem and the flow of information between thedifferent components and processes. The integration of the subsystem software andthe human operators or end-users involved is also considered.

42

Page 47: Conference Theme: Better Decisions for Africa

Determining an optimal vehicle fleet mix

Joke BuhrmannOPSI Systems, South [email protected]

Abstract

Transportation companies are often confronted with the question of how to selectan optimal fleet composition. Huge amounts of money can be wasted if the wrongvariety of vehicles is selected. For every situation, the fleet mix will differ. Somecompanies might have to select vehicles with large capacities to use on long tripsthat will prevent multiple trips to the same areas, while others might need to focusmore on obtaining small, flexible vehicles which may be used for multiple or adhoc deliveries during the day. The types of vehicles that the delivery points canaccommodate must also be taken into consideration. Practical points to considerwhen determining an optimal fleet mix for a company will be discussed in this paper.

43

Page 48: Conference Theme: Better Decisions for Africa

Developing spatial contiguity measures in multiobjectiveland use planning

Theodor J StewartUniversity of Cape Town, South [email protected]

Abstract

This presentation will review the development in thinking around decision supportsystem design for land use planning (largely based on collaborative work with theInstitute of Environmental Studies at the Free University of Amsterdam). Theproblems involve allocation of activities (e.g. agriculture, industry, conservation) todifferent parts of a region. Apart from the obvious criteria such as cost, conservation,agricultural outputs, there are less well-defined objectives of maintaining manage-able structures. It is typically not useful to have highly fragmented activities. Wewill review a number of approaches adopted, with the ultimate purpose of facili-tating group interactions in land use planning decisions. Initial approaches usedinteger linear programming with rather ad hoc spatial contiguity measures. Thiswas extended to a more formal problem structuring in terms of desired patterns ofclusters, and a genetic algorithm was designed, initially based on a grid structurebut later extended to parcels defined in a vector-based GIS. Although these workedwell, some recent simplifications will be discussed.

44

Page 49: Conference Theme: Better Decisions for Africa

Development and application of an assignment problem toreduce overcut waste in a secondary wood manufacturingfacility

Berndt G Lindner∗ & Tanya Lane-VisserStellenbosch University, South [email protected] & [email protected]

Abstract

Overcut waste costs in the wood manufacturing industry can be substantial, amount-ing to 30 m3 per month for the factory in question. By determining and assigningraw lengths more optimally this was reduced to 13.63 m3 through manual approxi-mation and further to 13.32 m3 by means of a mixed integer linear problem (MILP).Applying this operations research technique to the problem provided a solutionwhich reduced raw wood costs by R 50 130 per month. The model proves to workwell and may be used for determining optimal solutions in a shorter time and withfewer risks. Similar modelling development and application may be of great benefitfor the wood manufacturing industry.

45

Page 50: Conference Theme: Better Decisions for Africa

An econometric analysis of the effect of the ZimbabweStock Exchange on the Zimbabwean economy in itsyears of decline

Hausitoe Nare∗

National University of Science and Technology, [email protected]

Adrian T RamhewaPrivate Capacity, South [email protected]

Abstract

The Zimbabwean economy was in decline from the end of 1997 up to and includingthe year 2008. The economy experienced hyperinflation, a devaluing currency, ever-increasing money supply, soaring unemployment, closing down of companies, fallingproduction and shrinking real gross domestic product (GDP). In spite of the stateof the economy, the Zimbabwe Stock Exchange (ZSE) was recording positive returnson investment. It was even recognised as Africa’s best performing stock exchange atone point — yet the economy was Africa’s, as well as the World’s, fastest shrinkingeconomy outside of a warzone. This contradicts the universally accepted hypothesisthat the performance of the stock market is a barometer of economic activity in thecorresponding economy.

The focus of this paper is to establish the nature of the relationship between theZSE and the economy, that is, to determine whether the ZSE was damaging theeconomy, whether it had no effect on it at all, or whether it was itself a victim ofeconomic chaos. The study was done empirically by means of econometric modelsfor testing the ZSE industrial index against selected macroeconomic variables, suchas real GDP, broad money supply as well as the inflation rate. The results showthat the performance of the ZSE was actually driven by inflation with no evidencesuggesting that the ZSE industrial index, real GDP or broad money supply had anyeffect on each other.

46

Page 51: Conference Theme: Better Decisions for Africa

The effect of network structure and switching costs oninnovation diffusion in social networks

Thandulwazi Magadla∗ & Ian DurbachUniversity of Cape Town, South [email protected] & [email protected]

Abstract

The diffusion of innovations is influenced by various factors such as the manner inwhich agents are connected, interpersonal communication and product character-istics. This study presents a simple agent-based model for investigating the effectof various social network structures on the diffusion of innovations. We employ asatisficing framework in which agents use personal preferences as well as positiveand negative interpersonal communication to search for satisfying products. Wecompare the effectiveness of interpersonal communication in systems with switchingcosts and in systems without switching costs.

47

Page 52: Conference Theme: Better Decisions for Africa

Empirical results: Exact solutions using limited patterngeneration for the N-sheet Cutting Stock Problem

Tjaart Steyn∗ & Giel HattinghNorth-West University, South [email protected]

Abstract

Based on a theoretical framework, provable exact solutions for the two-dimensionalcutting stock problem for a specified number of same-size stock sheets using guil-lotine cuts in order to partially or totally satisfy a given order of demand itemsare considered. The main objective is to generalise the two-dimensional trim-lossproblem for a single sheet to N sheets.

Different algorithms were devised and implemented to evaluate the feasibility of thetheoretical framework. Empirical results are reported on a set of 120 instances basedon well-known problems from the literature. The results reported for this test bed ofproblems suggest this approach to be feasible for solving the cutting stock problemover more than one same-size stock sheet.

48

Page 53: Conference Theme: Better Decisions for Africa

Extreme daily increases in peak electricity demand:Tail-quantile estimation

Caston Sigauke∗

University of Limpopo, South [email protected]

Andrehette Verster & Delson ChikobvuUniversity of the Free State, South [email protected] & [email protected]

Abstract

A generalised Pareto distribution (GPD) is used to model extreme daily increasesin peak electricity demand. The model is fitted to 2000–2011 recorded data forSouth Africa in order to perform a comparative analysis with the generalised Pareto-type (GP-type) distribution. A Pareto quantile plot is used to obtain the optimumthreshold. Empirical results show that both the GP-type distribution and GPD area good fit to the data. One of the main advantages of the GP-type distribution is theestimation of only one parameter instead of two as is the case with GPD. Modellingof extreme daily increases in peak electricity demand improves the reliability of apower network if an accurate assessment of the level and frequency of future extremeload forecasts is carried out. One of the policy implications derived from this studyis the need for policy makers to design demand response strategies where users areexposed to electricity price-based incentives for them to save electricity particularlyduring periods of peak demand. Future peak electricity demand is influenced by thetails of probability distributions and not by means.

49

Page 54: Conference Theme: Better Decisions for Africa

The feasibility of a generalised additive neural networkfor spam classification

Pieter Labuschagne∗ & Tiny du ToitNorth-West University, South [email protected] & [email protected]

Abstract

The number of emails sent each year is increasing. People rely on emails as aconvenient way to communicate. Unfortunately, spammers abuse email services forcommercial propaganda. Currently more than 70% of all email traffic is spam. Thedrawbacks of spam left unattended can have significant financial consequences forcompanies. It can also pose a security risk and influence the overall performance ofa network. Internet service providers alone cannot fight this ongoing battle againstspammers. Researchers need to stay on the forefront of this problem so as to addressthe ever-changing tactics spammers use to bypass spam filters.

Multilayer perceptron (MLP) neural networks have been used successfully to identifyspam messages. Generalised additive neural networks (GANNs) are related to MLPsand have a number of favourable properties. This type of neural network has notbeen applied to spam filtering. In this study, the feasibility of employing a GANN todetect unwanted messages is investigated. GANNs overcome some of the problemsassociated with rule-based systems as they are capable of adapting to change withouthuman interaction. Insight into the relationships between inputs and the fittedmodel can be obtained by considering partial residual plots. This characteristic of aGANN helps overcome the black box effect, to which an a MLP, the most commonneural network, is prone. The performance of a GANN is compared to that ofa naive Bayesian approach and a memory-based classification technique using theLing-Spam corpus. It was found that the GANN outperforms these two classifierswhen flagging spam messages. This indicates that the feasibility of a GANN is worthinvestigating in more depth.

50

Page 55: Conference Theme: Better Decisions for Africa

Finding the best pass-receiving position in the RoboCupSmall-Size League

Moonyoung Yoon∗ & Tanya Lane-VisserStellenbosch University, South [email protected] & [email protected]

Abstract

As an attempt to provide a common platform for various research fields, the RoboCupoffers numerous problems for decision-making and optimisation. Specifically, theproblem of finding a good position for a pass-receiving robot is dealt with in thispaper. Besides the challenge of developing suitable evaluation criteria for assessingvarious field positions, another difficulty for approaches towards solving this problemis the time constraint. The solution should be found quickly enough to cope with alive field situation that is continually and rapidly changing.

In this paper a set of criteria for evaluating each position on the field and a CuckooSearch (CS) metaheuristic model are proposed to find the best field position fora pass-receiving robot. Furthermore, the run time of the algorithm is balancedbetween the quality of the solution and the agility in finding a solution. The solutionsfound by the CS algorithm are compared to explicit enumeration results for threegame situations. The comparison shows that the proposed metaheuristic model isapplicable in RoboCup Small-Size games.

51

Page 56: Conference Theme: Better Decisions for Africa

How healthy are the rhino populations in theHluhluwe-iMfolosi Park?

L Paul FattiUniversity of the Witwatersrand, South [email protected]

Abstract

Arising from a study conducted in the Hluhluwe-iMfolosi Park by the author inthe late nineties, a method is proposed for improving the estimate of the size of awildlife population by combining data from current and past surveys. The methodis based on a simple state space model which takes into account the (unknown) birthrate in the population and all known losses (mortalities and relocations) and gains(introductions) in the population between successive surveys, as well as the errors inthe survey estimates. The method is applied to the recent White- and Black Rhinopopulation estimates in the Hluhluwe-iMfolosi Park and tentative conclusions aredrawn on the health of these two populations.

52

Page 57: Conference Theme: Better Decisions for Africa

The hype around Analytics

Hans W IttmannHWI Consulting, South [email protected]

Abstract

Over the last few years there has been a lot of “hype” around analytics. For example,in the United States of America and the United Kingdom there is a huge drivearound this term, or concept, by operations researchers. Everything is referred to asAnalytics. Is this something that ORSSA should adopt more strongly? This paperwill endeavour to show the differences, if any, and the similarities between operationsresearch and analytics. In addition it will highlight the essence of what analyticsis and whether this is something that will enable operations research (under thebanner of analytics!) to be marketed “easier” and find wider acceptance. Variousapplications of analytics will be presented as well.

53

Page 58: Conference Theme: Better Decisions for Africa

An implementable routing solution for home-based carein South Africa

Nadia M ViljoenCouncil for Scientific & Industrial Research, South [email protected]

Abstract

Home-based care (HBC) is an effective service model for reducing the burden on acountry’s health and welfare systems. In South Africa, the orphaned and vulnerablechildren crisis has become the focus of HBC programmes. These programmes aremostly within semi-urban settlements where the need is greatest. Other successfulsolution approaches developed for HBC are not implementable in this low-technology,low-resource environment. A solution approach based on the space-filling curveheuristic is presented as an easily implementable, adequately performing alternativefor improving the routing of daily home visits.

54

Page 59: Conference Theme: Better Decisions for Africa

Improved techniques to model continuous operationswith discrete-event simulation

Anette van der MerweSasol Technology, South [email protected]

Abstract

Sasol, an integrated energy and chemicals company based in South Africa, leadsthe world in producing liquid fuels from natural gas and coal. Sasol faces manychallenges, such as stricter fuel specifications, fluctuating oil and gas prices as wellas complex and highly integrated production facilities. The company successfullyuses three discrete-event simulation models spanning its unique coal-to-liquids valuechain in order to improve decision-making. One of these models is the liquid factorymodel of Sasol’s synthetic oil refinery that is used to analyse the impact of majorinitiatives, test new operating philosophies and assist in removing production bot-tlenecks. The refinery, however, forms part of a larger value chain which includes thetar and diesel units that have not been included in the existing simulation model.These previously excluded units have now also been developed and incorporated intothe liquid factory simulation model using a stage gate model developed by Sasol’sDecision Support group. This presentation is a discussion on the challenges facedwhen simulating continuous operations using discrete-event simulation and includesa description of improved modelling techniques developed for modelling continu-ous operations. The stage gate model developed to facilitate successful modellingprojects is also discussed.

55

Page 60: Conference Theme: Better Decisions for Africa

Improving the work rate of community health workersthrough optimisation

Frans J Snyders∗ & Tanya Lane-VisserStellenbosch University, South [email protected] & [email protected]

Abstract

The aim of this paper is to show that an optimised route can reduce the travel timeand cost required for community health workers (CHWs) to perform their work andto get a sense for the magnitude of potential saving when applied to a realistic casestudy. Through optimised route planning CHWs can achieve higher work rates,resulting in more effective resource utilisation. One of the tasks of CHWs is captur-ing health-related data. Technology exists that improve the efficiency of capturingdata through using mobile phones (such as EpiSurveyor). The problem lies in thatCHWs have to visit many locations to capture data across large distances, causingCHWs to spend a lot of their time travelling. In this paper the CHWs situation ismodelled as a travelling salesman problem (TSP). The paper shows how optimisedroute planning can be used by CHWs to reduce their time spent on travelling andreduce the cost required to perform their work.

56

Page 61: Conference Theme: Better Decisions for Africa

Integrating technology into teaching and learning in theschool classroom

Olatunde O OsiyemiUniversity of Fort Hare, South [email protected]

Abstract

Twenty-five years ago, the term technology had a rather different meaning than itdoes today. Our education processes often reflect somewhat conservative ideas ofwhat learning is and how it should be taught. The changing nature of the secondaryschool system is an ongoing challenge, often demanding quite radical reforms ineducation.

Information is needed on the best ways to integrate technology into school classroomsettings, as well as how to best prepare teachers to use technology. The integrationof learning technologies into high school classrooms is being promoted and supportedaround the world. Underlying the promotion and support are claims that successfulintegration will lead to enhanced learning outcomes.

The phenomenographic research approach will be used to analyse some of the dis-tinctly different ways that learning technologies can be perceived by teachers andlearners in classrooms. Are the perceptions consistent with the integration of tech-nology in teaching and learning in classrooms in a manner likely to encourage en-hanced learning outcomes? The results from the use of learning technology willenhance learning outcomes; better learning techniques and strategies; skills in usingthe technology; assistance and motivation from the technology; and more effectivepresentation of the learning outcomes. Learners’ approaches to learning are relatedto their teachers’ approaches to teaching.

57

Page 62: Conference Theme: Better Decisions for Africa

An intelligent particle swarm optimisation frameworkbased on a multi-agent system

Nachamanda V Blamah∗ & Aderemi O AdewumiUniversity of KwaZulu-Natal, South [email protected] & [email protected]

Gregory M WajigaModibbo Adama University of Technology, [email protected]

Benson Y BahaInformation Technology & Systems, Mainstreet Bank, [email protected]

Abstract

Particle swarm optimisation techniques typically comprise a population of simpleagents interacting locally with one another and with their environment, with thegoal of locating optima within the operational environment. In this paper, a robustand intelligent particle swarm optimisation framework based on multi-agent systemis presented, where learning capabilities are incorporated into the particle agents soas to dynamically adjust their optimality behaviours. Autonomy is achieved by theuse of communicators that separate an agent’s individual operation from that of theswarm, thereby making the system more robust.

58

Page 63: Conference Theme: Better Decisions for Africa

An inventory routing system for dynamic food distribution

Hildah MashiraNational University of Science and Technology, [email protected]

Abstract

This presentation is concerned with designing a routing system for improving fleetutilisation and reducing delivery costs whilst satisfying customer demand as wellas maintaining optimum inventory levels for a fast food distribution centre. Thestudy centred on Harare’s distribution for this company. An algorithm, called theA*-algorithm, was adopted to solve the problem. Because the problem is a combi-natorial problem, simulation software was developed to solve the problem.

59

Page 64: Conference Theme: Better Decisions for Africa

Keeping it simple in a data-sparse environment: The caseof donor breastmilk demand and supply in South Africa

Nadia M Viljoen∗

Council for Scientific & Industrial Research, South [email protected]

Julie Swann, Melih Celik, Wenwei Cao & Ozlem ErgunGeorgia Institute of Technology, United States of [email protected], [email protected],[email protected] & [email protected]

Abstract

Donor breastmilk may potentially save thousands of neonatal lives and save millionsof Rands in treatment costs annually. A facility location-allocation model will beused to develop a strategic national network expansion plan based on an existingbreastmilk banking service model. The disaggregate demand and supply data re-quired by this location-allocation model does not exist as-is in South Africa. This isoften the case when developing operations research models for developing countries.The case study illustrates a simple methodology that combines demographic data,health statistics and insights from the literature and subject experts to determinethat in 2011 almost 90 000 premature infants without access to mother’s-own-milkwould have required more than 1.7 million bottles of pasteurised donor breastmilkto protect them from fatal infections during the first 14 days of their lives. Simul-taneously, 160 000 bottles of unpasteurised donor breastmilk could be sourced frompotential donors. The disaggregate estimates show that supply and demand aregeographically disparate and that at most 43% of demand could be covered withthe given demand. This has implications for the model development, specifically inaccounting for equitable distribution.

60

Page 65: Conference Theme: Better Decisions for Africa

A linear response surface analysis approach to evaluateQoS in wireless networks

Andre Brand∗, Hennie Kruger & Henry FouldsNorth-West University, South [email protected], [email protected] & [email protected]

Abstract

The growth of wireless networks and the increase in personal internet use for diverseapplications has made it important to deliver good quality of service (QoS) to usersof these wireless networks. This study proposes a linear response surface analysisapproach towards evaluating QoS factors such as throughput and delay. For thepurpose of this study a 802.11n prototype wireless network was constructed usingdifferent configurations. Multiple audio, text and ZIP files were sent over the net-work and QoS data, such as throughput, delay, jitter, packets sent/received and thenumber of nodes, were recorded for each of these transmissions. The objective wasto construct a linear response surface analysis model with which the data can beevaluated. The model allows for graphical comparison of the measured data pointsin order to evaluate the impact they may have on the QoS provided by the network.

61

Page 66: Conference Theme: Better Decisions for Africa

The management of investment portfolios usingoptimisation models

Brian MudharaNational University of Science and Technology, [email protected]

Abstract

Investors and investment institutions all over the world are faced with the challengeof determining worthwhile investments. They seek to maximise their wealth and toavoid making losses in the process. To do this investors pool different investmentstogether by which they expect to make profit or gain income; this is called coming upwith an investment portfolio. This project seeks to aid an investment institution incoming up with a proper investment portfolio that maximises returns and minimisesrisks of loss. The main method that is employed is quadratic programming. Otherfinancial mathematics techniques and theories, such as the mean variance theory,will be incorporated to this project. Management Scientist and Matlab are used tosolve the formulated mathematical model.

62

Page 67: Conference Theme: Better Decisions for Africa

Meta modelling with linear programming for capitalinvestment with uncertain returns

Dave W EvansDevelopment Bank of Southern Africa, South [email protected]

Abstract

The use of multi-time period linear programming models to evaluate capital invest-ments is common in the petro-chemical industry. The model is typically set up toinclude the new capacity being considered, over a reasonable time scale — typicallyten to fifteen years. Cases are then run with and without the capacity available,and the additional profit from the new capacity is used in conjunction with theexpected initial capital cost, to see whether an acceptable internal rate of return isachieved. This paper considers a higher level of modelling, used in conjunction withthis approach, needed in an unusual situation where it was virtually impossible topredict the profit expected from the capital investment.

63

Page 68: Conference Theme: Better Decisions for Africa

A model and results of local search heuristics for theHostel Space Allocation Problem

Sunday A Ajibola & Aderemi O Adewumi∗

University of KwaZulu-Natal, South [email protected] & [email protected]

Abstract

The optimisation of scarce beds space allocation to large numbers of students resultedin the hostel space allocation problem (HSAP). In this paper two local search heuris-tics, namely hill climbing and tabu search, are investigated for solving the HSAP.We propose new models to solve the multi-staged, multi-constrained and multi-objective HSAP, including the floor allocation stage, which is seldom treated in theliterature due to its complexity. Results obtained by means of the two algorithmsare compared with solutions available in the literature for this instance of the SpaceAllocation Problem (SAP).

64

Page 69: Conference Theme: Better Decisions for Africa

Modelling the impact and cost of the sterile insect techniqueon Eldana saccharina Walker in sugarcane

Linke Potgieter∗ & Jan H van VuurenStellenbosch University, South [email protected] & [email protected]

Abstract

A mathematical model is formulated and analysed for an Eldana saccharina infes-tation of sugarcane under the influence of partially sterile released insects. Themodel describes the population growth and interaction between normal and sterileE.saccharina moths in a temporally variable environment. The model consists of adeterministic system of difference equations. The primary objectives is to determinesuitable parameters in terms of which the population growth and interaction maybe quantified and according to which infestation levels and the associated sugar-cane damage may be measured. Using the model, the efficiency of different sterilemoth release strategies are investigated without having to conduct formal field ex-periments, and guidelines are presented according to which release ratios, releasefrequencies and spatial distributions of releases may be estimated which are ex-pected to lead to suppression of the pest. The model is the first to describe thesterile insect technique applied specifically to E.saccharina growth, and to considerthe economic viability of applying the technique to this species.

65

Page 70: Conference Theme: Better Decisions for Africa

Modelling and optimisation of EOQ in supermarkets understochastic demand: A case study involving a milk powderproduct in Uganda

Kizito P Mubiru∗

Kyambogo University, [email protected]

Kariko B Buhwezi & Peter LatingMakerere University, [email protected] & [email protected]

Abstract

Supermarkets continually face the challenge of optimising the economic orderquantity (EOQ) of cycle inventories in a stochastic demand environment. In thispaper, a mathematical model is proposed for optimising the EOQ and inventorycosts of milk powder under stochastic demand. Adopting a Markov decision processapproach, the states of a Markov chain represent possible states of demand for a milkpowder product. The decision of how much to order is made using dynamic pro-gramming over a finite planning horizon. The approach demonstrates the existenceof an optimal state-dependent EOQ as well as the corresponding total inventorycosts.

66

Page 71: Conference Theme: Better Decisions for Africa

Modelling and optimisation of production lot sizingdecisions under stochastic demand: A case study involvinga milk powder product in Uganda

Kizito P Mubiru∗

Kyambogo University, [email protected]

Kariko B Buhwezi & Peter LatingMakerere University, [email protected] & [email protected]

Abstract

Effective production-inventory management requires cost-effective methods fordeciding on optimal lot sizes of cycle inventories in a stochastic demand environment.In this paper, we develop a mathematical model for optimal lot sizing decisions,production and inventory costs of milk powder given a periodic review production-inventory system under stochastic demand. Adopting a Markov decision processapproach, the states of a Markov chain represent possible states of demand for amilk powder product. The decision of when to produce additional units is madeusing dynamic programming over a multi-period planning horizon. The approachdemonstrates the existence of an optimal state-dependent lot sizing decision as wellas the corresponding total production and inventory costs.

67

Page 72: Conference Theme: Better Decisions for Africa

An operations management algorithm in the tilemanufacturing industry

David C ZviporeNational University of Science and Technology, [email protected]

Abstract

The tile manufacturing industry is a fundamental production sectors in which,despite increasing demand for tiles and increasing capacity of the major players,the majority of Zimbabwean companies have failed to fully exploit high demandlevels due to infrequent breakdown of machinery, poor production management andcapacity planning.

This research paper aims to combine optimisation algorithms, capacity planning,scheduling and discreet event simulation of a tile manufacturing plant, by determin-ing the tooling, personnel and equipment resources that are required for optimalefficiency in the manufacturing process. A Dynamic programming knapsack algo-rithm is used to optimally select jobs (in a twelve-stage manufacturing process) suchthat they contribute to the production plan within a specified duration time. Thisis done in such a way that each stage contributes to an overall optimal productionplan for the tile manufacturing process at minimal costs.

An Arena Simulation model is used to validate the manufacturing process design andidentify areas of process improvement. Recommendations include the design of pro-duction processes to match volume-variety requirements, process design positioningand the incorporation of appropriate process technology. Furthermore, job designs,recognising process variability, appropriately configuring process tasks, capacity andadopting a minimal cost task-precedence model have also been found by means ofthe model.

68

Page 73: Conference Theme: Better Decisions for Africa

Operational research in the forestry industry — Focussingon a recently developed allocation model

Margarete J [email protected]

Abstract

Throughout the world the forestry industry daily faces logistical challenges anddecisions. Operations research has been playing a significant part in this industryfor years.

This talk will focus on the implementation of a mathematical programming modelin AIMMS that assists foresters in South Africa. The purpose of this model is toensure that timber is allocated to customers in an optimal manner according to anobjective and fair bidding process.

A further problem of ensuring a sustainable timber supply to customers, by schedul-ing harvesting teams optimally throughout the financial year, was solved in anotherAIMMS model. This model will also be discussed briefly.

69

Page 74: Conference Theme: Better Decisions for Africa

An optimal distribution strategy for a supply chainorganisation: The case of Delta Beverages

Absalom Jaison∗ & Desmond Mwembe∗

National University of Science and Technology, [email protected] & [email protected]

Justin Chirima∗

Great Zimbabwe University, [email protected]

Abstract

Supply chain management is a strategic management operation that is involvedwith the movement of goods from suppliers to factories to warehouses to retailersand finally to customers. Organisations need to manage this as to reduce unnec-essary costs. Delta Beverages uses this management operation in the productionand distribution of their commodities. A best operation strategy will reduce distri-bution costs as much as possible. In this research customers are considered whosedemand for soft drinks is 100 crates and above per month from Bulawayo DeltaBeverages. Customer locations are put in a Cartesian (x, y)-plane for easy location.Their weight (distances and demand) and distribution costs from the current singlewarehouse were calculated and these are compared with those obtained from thelocations determined using clustering. A financial implication of having an optimalnumber of warehouses is analysed. Using the nearest centre reclassification algo-rithm together with Weisfeld’s algorithm, two, three and four cluster (warehouses)are considered and calculations for distribution costs for each customer is done. Asimulation of this strategy is performed in Arena considering randomly selected cus-tomers from each cluster. Results are compared with those obtained from MicrosoftExcel results. A best operation strategy is recommended for Delta Beverages. Aproject valuation of either to expand or remain with a single warehouse is doneusing the real option valuation technique and a recommendation is done.

70

Page 75: Conference Theme: Better Decisions for Africa

The optimisation of terrain modelling and line-of-sighttechniques

Andries M Heyns∗ & Jan H van VuurenStellenbosch University, South [email protected] & [email protected]

Abstract

Applications performed upon computerised representations of geographical terrainmodels vary according to radar placement in military strategic planning or towerplacement in telecommunication applications. The run time of these applicationsand the accuracy of their results — which may include specific coordinates andvisibility computations — depend on the accuracy and reliability of the terrain dataupon which they are performed. It is therefore necessary to model the terrain so asto include such features as the curvature of the earth and to consider the effects oflatitudinal and longitudinal variances due to the shape of the earth.

Once an acceptable terrain model has been generated, various line-of-sight com-putations may be performed thereupon and the resulting running-times of thesecomputations by different techniques may be compared to each other. Differentline-of-sight techniques yield different performance results, depending on the type ofterrain (e.g. smooth or rough). It is therefore possible to design a hybrid techniquewhich exploits the advantages of different line-of-sight techniques for the type ofterrain to which it is suited.

71

Page 76: Conference Theme: Better Decisions for Africa

Optimised schedules achieve massive savings

Robert A BennettoOPSI Systems, South [email protected]

Abstract

The rail network supplying coal from Mpumalanga to Richard’s Bay delivers 67million tonnes and earns billions of Rands per annum. This crucial resource hasmore than 7 000 dedicated wagons and 160 locomotives with a combined value ofroughly 13 billion Rands. Planning all of these resources is a huge and complicatedchallenge, and even the smallest reduction in waiting times translates into millionsof Rands saved.

Working with Transnet Freight Rail, we have developed and implemented a unique,dynamic scheduling system that uses a genetic algorithm to find a good solution andthen is able to re-plan at any time based on the current positions of the resources,track availability and mine demand. In this talk we will discuss how the train line ismodelled, the complex types of train moves available, the rich logistical constraints,how the genetic algorithm works, and the benefits that have been achieved. Abrief introduction to scheduling problems in general and an explanation around whyNP-completeness makes these problems so hard will also be given.

72

Page 77: Conference Theme: Better Decisions for Africa

Practical application of semi-supervised segmentation withina predictive modelling context

Gerbrand D Breed∗, Tanja de la Rey & Stephanus E TerblancheNorth-West University, South [email protected], [email protected]& [email protected]

Abstract

Predictive models are widely used to optimise and improve processes. Industrystandards and best practices on robust model development have been refined overmany years, and even though many software tools are available to simplify theprocess today, developing a practically implementable model for long-term use stillinvolves substantial human intervention.

Segmentation of predictive models has been well-established in the industry as partof the model development process. Segmentation is performed with the aim ofimproving overall predictive performance, and is required when the characteristicsavailable to model the outcome differ in some way between segments. Current seg-mentation techniques can be split into two main streams: The first stream is basedon maximising target separation or impurity between segments through supervisedclassification, most commonly decision trees. The second stream defines segmentsby maximising the dissimilarity of the character distribution based on a distancefunction (for example, Euclidean distance in cluster analysis).

The first stream focuses on the dependent (target) variable, while the second streamfocuses on the distribution of the independent variables to be used for modelling.Both streams have been successful in improving model accuracy and, depending onthe application, have outperformed each other in different environments.

We propose a methodology which balances both the use of the target variable as wellas the distribution of the independent variables during segmentation. We elaborateon the detail behind the methodology, illustrating its advantages and disadvantages,and apply it to industry data in order to compare the accuracy of the segmentedmodels developed to the results obtained from the standard methodologies.

73

Page 78: Conference Theme: Better Decisions for Africa

A practical application of stochastic simulation within anexplosives supply chain environment

Jaco L Joubert∗ & Gerhard CarstensSasol Shared Services, South [email protected]

Abstract

The Sasol nitro explosives business unit launched a supply chain improvementproject and identified finished goods storage capacity as a constraint. It is, how-ever, a complicated and long process to expand storage capacity due to safety /EIA regulations and therefore imperative that available capacity be used optimally.Stochastic simulation was used to demonstrate the benefits of maximising prod-uct flow across the supply chain rather than optimising each area separately. Analternative inventory management system resulted in an increase in supply chainthroughput capacity without additional storage capacity.

74

Page 79: Conference Theme: Better Decisions for Africa

A problem reduction approach for the Survivable NetworkDesign Problem

Stephanus E TerblancheNorth-West University, South [email protected]

Abstract

A traffic demand matrix defines the traffic requirements of a capacitated networkdesign problem. More specifically, it describes the traffic load between each origin-destination pair of nodes. Topology design and capacity placement depend on theinformation provided by these matrices. It is, however, unlikely that a single trafficdemand matrix is sufficient for describing varying traffic conditions. Solving thestochastic or robust network design problem therefore takes a large number of mul-tiple traffic demand matrices as input. The resulting network design should be ableto support all of these matrices non-simultaneously. Traffic data can be measuredfrom operational networks or generated through simulations resulting in thousandsof multiple traffic demand matrices.

An approach will be presented for identifying the dominant set of traffic demandmatrices by taking survivability into account. The matrices being dominated by thedominant set can be excluded from the network design problem without altering thesolution space, resulting in reduced computing times.

75

Page 80: Conference Theme: Better Decisions for Africa

Replication of base models to improve the performanceof positive vs negative classification ensembles forpredictive data mining

Patricia EN LutuUniversity of Pretoria, South [email protected]

Abstract

The creation of classification models is a common activity in predictive data mining.A classification ensemble consists of many base models (classifiers) whose predictionsare combined into one prediction using a combination algorithm. Most commonly,each base model in the ensemble is capable of predicting any of the classes for theprediction task. Positive versus negative (pVn) classification has recently been pro-posed in the literature as an ensemble classification method with a potential toprovide high predictive performance for multiclass prediction tasks. The studiesreported to date on pVn classification ensembles were based on the use of one basemodel for each group of classes. The purpose of this paper is to report on the exper-imental results for the predictive classification performance of pVn ensembles wherethe base models that predict minority classes are replicated. The KDD Cup 1999data set, available from the UCI KDD archive, and the multilayer perceptron artifi-cial neural network (MLP ANN) classification algorithm were used for the reportedexperiments. It is demonstrated in this paper that the use of pVn ensemble classifi-cation can provide significant improvements in predictive performance for datasetswith minority classes when the base models for the minority classes are replicated.

76

Page 81: Conference Theme: Better Decisions for Africa

The Residential Waste Collection Problem in an urban areaunder constraints of restricted public infrastructure

Kala KJ Robert∗ & Mourdoun T ChristelleCatholic University of Central Africa, [email protected] & [email protected]

Abstract

Due to the sharp increase of urban populations and restricted public infrastructure,waste collection has become a major concern for municipality leaders in developingcountries. This study is concerned with the efficient management of logistics forresidential waste collection in Douala city in Cameroon, which has a populationof more than 2 million people. More precisely, a model is developed to optimisethe tour of one vehicle in a solid waste door-to-door collection system. The cargocollected by the vehicle is unloaded in a single dumpsite. Furthermore, the goal isto reduce both the financial cost of the route followed by a vehicle and to minimiseits time spent on this route. The model is based on three main concepts:

(1) A weighted graph of the routes constructed from a geo-referenced map,

(2) A method for determining arcs’ weightings based on operational data, thequality of public infrastructure and management accounting, and

(3) A heuristic to solve the capacitated arc routing problem (CARP) based on theshortest path problem on a graph topology.

As described in the contract between the waste treatment company and the munici-pality of Douala, the model takes into account the quantity of waste to be collectedper day and the area to be covered.

77

Page 82: Conference Theme: Better Decisions for Africa

Results of local search heuristics for the Annual CropPlanning Problem

Sivashan Chetty∗ & Aderemi O AdewumiUniversity of KwaZulu-Natal, South [email protected] & [email protected]

Abstract

This paper presents the results of three local search heuristics for the NP-hardAnnual Crop Planning (ACP) problem. A new local search technique, called aCompetitive Algorithm (CA), was designed and implemented for this purpose. Theresults obtained from the CA are compared with those obtained by known localsearch techniques such, as tabu search and simulated annealing.

78

Page 83: Conference Theme: Better Decisions for Africa

The science and magic of scheduling

David J LubinskyOPSI Systems, South [email protected]

Abstract

In business doing things right is the first goal, but doing things optimally takesoperations to a new level. With proper planning the same resources can achievebetter productivity, better customer service and much better visibility. Many peopledo not realise that simply having a good repeatable process might actually be costingthem money by not planning, but on the flip-side is that implementing advancedplanning systems in the transport arena can be difficult and often such projectsfail. In this talk I will introduce some of the science of scheduling algorithms anddiscuss a few case studies from small single depot sites to large enterprise roll-outsof routing and scheduling systems. I will also discuss the scheduling system that webuilt for the Transnet coal line which has helped them to realise billions of Randsin increased productivity. These case studies will emphasise the do’s and the don’tsof implementing this kind of advanced software to give some advice on how to avoidyet another expensive software failure.

79

Page 84: Conference Theme: Better Decisions for Africa

Self-organising traffic light control

Mark D Einhorn∗, Jan H van Vuuren & Alewyn P BurgerStellenbosch University, South [email protected], [email protected] & [email protected]

Abstract

Traffic congestion is a major problem in urban areas around the world, resulting insignificant costs in terms of man hours lost and excessive fuel consumption. Certainsteps have been taken in various cities in an attempt to ease traffic congestion andthereby lessen its debilitating effects. These steps include the widening of roadways,the introduction of inner-city congestion charges, improvements to public transportsystems and improved traffic control strategies at road intersections, to name a few.

In this paper, a traffic control technique is presented which utilises self-organisingalgorithms to determine the switching sequences of traffic control signals at an inter-section. The input data to the algorithms are provided by radar detection equipmentmounted at the intersection, allowing the traffic lights to monitor a length of road-way and determine the most effective switching strategy based on the speeds ofthe approaching vehicles as well as their respective distances from the intersection.The algorithms attempt to incorporate vehicle clustering heuristics, with variousclustering criteria being presented.

The self-organising traffic control algorithms are tested in a simulated environmentusing a traffic simulation model designed and built specifically for the purpose of thestudy. Results are shown for a number of varying road network topologies and trafficflow profiles. The results are compared to those obtained by a fixed time regime forwhich the green times and off-sets among adjacent intersections have been optimisedin terms of average driver waiting times, average time spent by vehicles in the systemand the total average queue lengths for the system.

80

Page 85: Conference Theme: Better Decisions for Africa

Solving a binary linear programming model in polynomialtime

Elias MunapoUniversity of South Africa, South [email protected]

Abstract

A technique is presented for solving a binary linear programming (BLP) model inpolynomial time. A combination of some duality and complementary slacknesstheories are used to linearise the BLP. The optimal solution of the resuling LP isexactly the optimal solution of the original problem.

81

Page 86: Conference Theme: Better Decisions for Africa

Spreadsheet solvers: A comparison

Wim R GeversStellenbosch University, South [email protected]

Abstract

The Solver that is distributed by Microsoft as part of Excel is probably well knownto persons using spreadsheet solvers. This software is limited to small problems. Toupgrade one can acquire the Risk Solver Platform from Frontline Systems, but thisis not cheap. Recently an open source spreadsheet solver became available. Thispresentation will highlight some of the more important features of the open sourcesolution and compare this to commercial software solutions.

82

Page 87: Conference Theme: Better Decisions for Africa

Stochastic simulation for Sasol Solvents Global TankPlanning

Ester J Vermaak∗, Leilani E Meijer & Louis H SnydersSasol Technology, South [email protected], [email protected]& [email protected]

Abstract

The stochastic Sasol Solvents Global Tank Planning model was developed to providea dynamic view of the global storage network for the Solvents business unit. Thismodel provides the capabilities to simulate and evaluate current and future storagecapacities and inventory development for multiple products when considering volumeallocations and sales patterns in different regions. It covers seven sales regions fromthe Americas to Europe, the Middle and Far East. The simulation model has beenused to assist in the business case for a re-configuration of current storage capacitiesand can be used in future to support various Sasol Solvents supply chain networkdecisions.

83

Page 88: Conference Theme: Better Decisions for Africa

Strategic coordination in a production supply chain

Tichoana C MazuruNational University of Science and Technology, [email protected]

Abstract

This paper considers how production time and labour can be utilised to producea maximum output at minimum levels in production supply chain. We investigatethe feasibility of using linear programming and simulation in managing productiontime and labour required. The research entailed developing and using linear pro-gramming to come up with lead time and labour requirements. A simulation modelwas developed which mimics plant operations whilst using the input results fromthe linear program. Through experimentation using the Tora evaluation softwareand the Arena simulation environment, significant (up to 12.5%) savings on labourwere achieved. The production lead time was also reduced by 3%. These savingsare notwithstanding the implementation of effective supply chain and productionmaintenance.

84

Page 89: Conference Theme: Better Decisions for Africa

A statistical analysis of monthly temperature usingBox-Jenkins’ Arima methodology and a general linearmodel approach: A case study of the city of Bulawayo

Farikayi K Mutasa∗, Nonhlanhla Magadlela,Philimon Nyamugure & Edward T ChiyakaNational University of Science and Technology, [email protected], [email protected],[email protected] & [email protected]

Abstract

Climate change has been widely blamed to have caused an increase in unstableweather events such as high temperatures. Over the past centuries, climate changehas had a great influence on natural ecosystems and social economics. As a resultstudies on temperature have become increasingly important in recent years. Tem-peratures have been recorded in Bulawayo for a long time. The main aim of thispaper is to analyse Bulawayo monthly temperature data by statistical methods,and to build a general linear model (GLM) and an autoregressive integrated mov-ing average (ARIMA) model to fit the data for a period of 30 years, from 1977 to2007. Forecasting models for the mean monthly temperature are developed and thenused to forecast the monthly temperature for 2008 and a comparison is made withthe actual mean monthly values. After comparing the results the seasonal ARIMAmodel for the series was found to fit the data better than the GLM.

85

Page 90: Conference Theme: Better Decisions for Africa

Studies in metaheuristics for the Blood AssignmentProblem

Emmanuel Dufourq, Michael O Olusanya∗ & Aderemi O AdewumiUniversity of KwaZulu-Natal, South [email protected], soji [email protected] & [email protected]

Abstract

Every single day there is demand for blood transfusion. Units of blood do nothave an unlimited shelf life and hence must be assigned to patients rapidly. Bloodbanks continually need to meet the demand for blood but when when inventoriesare depleted they have to import blood from outside the system. This study is acomparison of different heuristics which have been applied to the Blood AssignmentProblem. These include genetic algorithms, hill climbing and simulated annealing.Several mutation operators were designed. Genetic algorithms were found to yieldthe best results overall.

86

Page 91: Conference Theme: Better Decisions for Africa

A successful case study for integration between business andanalytics in the Propylene value chain

Nica van der WesthuizenSasol Shared Services, South [email protected]

Abstract

Molecules meant for chemical manufacturing were lost along a very integrated anddynamic value chain serving multiple businesses of which the Petrochemical refineryand the Polymers businesses are a major part. Sound engineering principles were in-corporated in an Excel based mass-balance tool which was used hand in hand with anintegrated simulation model of the propylene value chain. The combined analyticscapability resulting from this integrated approach were able to show additional stor-age requirements and identify pinch points and constrains. The tested scenariosunpacked the interactive effects of the value chain under different conditions andincreased the integrated businesses understanding. This also led to the creation ofa road map for implementation of various projects over the next ten years.

87

Page 92: Conference Theme: Better Decisions for Africa

A supervised learning approach for file fragmentclassification

Erich Wilgenbus∗, Hennie Kruger & Tiny du ToitNorth-West University, South [email protected], [email protected] & [email protected]

Abstract

The file block classification problem is of importance in data recovery, memoryanalysis as well as network and computer forensics. This classification problem re-quires prediction of the correct file type to which a given file fragment belongs.Several approaches have been proposed in the literature. This research shows howa neural network and linear programming discriminant model can be used togetherto solve pairwise binary classification problems. The multi-class problem is solvedusing a sequence of these pairwise binary classifiers. The results of empirical exper-imentation will be presented.

88

Page 93: Conference Theme: Better Decisions for Africa

A survey of hyper-heuristics for the Nurse RosteringProblem

Nelishia Pillay∗ & Christopher Rae∗

University of KwaZulu-Natal, South [email protected] & [email protected]

Abstract

The First International Nurse Rostering Competition held in 2010 has stimulatednew interest in this domain. Various techniques, such as tabu search, geneticalgorithms, simulated annealing, integer programming and constraint programming,have been applied to solving the nurse rostering problem. Hyper-heuristics, afairly new approach, searches a heuristic space rather than a solution space in anattempt to provide a more generalised solution to a problem. There are essentiallyfour categories of hyper-heuristics, namely, selection constructive, selection pertur-bative, generation constructive and generation perturbative. This paper provides anoverview of the use of hyper-heuristics for solving the nurse rostering problem and,based on a critical analysis of the application of hyper-heuristics to this domain,proposes directions for future research.

89

Page 94: Conference Theme: Better Decisions for Africa

Teachers’ and learners’ perception on mathematics literacyas subject stream

Olatunde O OsiyemiUniversity of Fort Hare, South [email protected]

Abstract

The research will discuss teachers’ and learners’ perception on mathematics literacyas subject stream in the final three years (grade 10 to 12) of schooling in SouthAfrica. Subject streaming reduces the personality and confidence of learners whichalso includes: attitude, interest areas, low esteem and skills. Research questions willbe what teachers’ and learners’ perceptions are towards mathematical literacy? Theteachers and learners perception on mathematical literacy as subject stream in thefurther education and training phase will be investigated using mixed methods.

90

Page 95: Conference Theme: Better Decisions for Africa

Using cluster analysis to extract trip and activityinformation from GPS data

Johanna H Nel∗ & Stephan C KrygsmanStellenbosch University, South [email protected] & [email protected]

Abstract

Travel surveys which include, for example, questions on mode of travel, length oftravel, type of activities and the length of activities are essential requirements fortransport planning. Conventional travel diaries and questionnaires are often tediousto complete and result in missing information.

While still a relatively under-researched field, the use of Global Positioning Systems(GPS) to collect travel behaviour data has already made some contributions withrespect to the accuracy of trips and activity reporting. It also provides many benefitsfor respondents such as low input requirements and thus little (if any) respondentburden (such as trying to recall activities or trips).

However, few studies report actual comparisons between analysed GPS data andtravel surveys with reference to travel time, number of trips, number of activities,the nature of the activities and distance travelled. Most of the studies incorporateGIS systems and land use maps in their analyses. Proper land use maps are notalways available in developing countries.

The overall objective of this research was to develop a new approach towards deter-mining the usefulness of GPS in travel and activity surveys, without the availabilityof land-use maps. A total of 176 employees at a university took part in the GPStracking study. They were tracked for two days, and completed a travel diary forthose two days as well as a household travel questionnaire. A new method, employ-ing cluster analysis, was developed to convert the GPS records of each participantinto an activity diary, as well as a travel diary. Trip statistics and activity statis-tics were compared with actual values extracted from the travel diary. The resultscompare favourably with those from the travel diary.

91

Page 96: Conference Theme: Better Decisions for Africa

Case Study Tutorial:Using operations research for strategic planning in thenot-for-profit health sector — A case study

Nadia M Viljoen∗

Council for Scientific and Industrial Research, South [email protected]

Wenwei Cao, Melih Celik, Julie Swann & Ozlem ErgunGeorgia Institute of Technology, United States of [email protected], [email protected],[email protected] & [email protected]

Abstract

Globally, 4 million newborn babies die annually in the first 28 days of their lives —neonatal infections being among the leading causes of death. Breastfeeding is themost effective way to reduce these infections, but in South Africa many infants areborn without access to Mother’s-own-Milk (MoM) due to poor maternal health andthe lack of rooming-in facilities in public hospitals. Breastmilk donation services inSouth Africa have protected thousands of premature infants from the risk of fatalinfection over the last few years.

This case study explores how OR can assist in planning the structure of a nationalcollection and distribution network. The first module of the study discusses theunique humanitarian context to be captured in the OR model. The second modulepresents the data collection and processing, illustrating the problem’s sensitivity todata availability and disaggregation through an interactive spreadsheet tool. Thethird module allows the participant to compete interactively with the optimisationmodel on a small data set to try and find the most equitable network design in theCape Town area.

92

Page 97: Conference Theme: Better Decisions for Africa

Using multi-criteria decision aid in corporate climate changeresponse

Muriel Chinoda∗ & Jan W KrugerUniversity of South Africa, South [email protected] & [email protected]

Abstract

The heightened risks and opportunities posed by climate change call for increasedattention by business executives to employ creativity and rigorous methodology ingenerating strategic response options. Because of the diversity of climate changeresponse initiatives and activities, companies are not expected to be currently usingall possible climate change response strategies. It is most likely that each company isprogressing on its green journey by adopting those strategies and initiatives whichthey understand are valuable and/or have merit based upon some criteria suchas previous experience skills sets, immediate future threats, vulnerabilities, risks,opportunities and other drivers. The analytical hierarchy process is used as anappropriate multi-criteria decision support tool that allows executives to have ade-quate regard for: the interconnectedness of climate change systems, the constraintsin time, knowledge, computational abilities that humans face, and the irrationalmanner in which humans make decisions.

93

Page 98: Conference Theme: Better Decisions for Africa

The weapon assignment scheduling problem in a ground-based air defence environment

Michelle van der Merwe∗ & Jan H van VuurenStellenbosch University, South [email protected] & [email protected]

Abstract

A Threat Evaluation and Weapon Assignment (TEWA) system is a decision supportsystem designed to evaluate threats in real-time, coupled with suggesting weapon-threat assignments in order to destroy or scare away enemy forces attempting toattack Defended Assets (DAs). In a Ground-based Air Defence (GBAD) environmentthere are typically a number of DAs on the ground which require protection againstaerial threats. A number of radar sensors are employed to search for and trackpotential aerial threats, and various Weapon Systems (WSs) are used to engage thethreats. The aim in this talk is to focus on the Weapon Assignment SchedulingProblem (WASP) in a GBAD environment — a subproblem which must be solvedwithin a TEWA decision support system.

The WASP consists of assigning a set of available WSs to a set of aerial threats, aswell as scheduling time intervals during which engagements should occur. The aimis to minimise the total expected survival value of all the threats that survive anengagement by the WSs. The proposed assignments are sent to a fire control officerwho is responsible for actually assigning WSs to targets, in order to aid him whenexecuting fire control.

The WASP is formulated as a vehicle routing problem with heterogeneous vehicleswhere each customer has to be served within a certain time window by some vehicle.Two metaheuristics, namely simulated annealing (SA) and a hybrid between SAand a tabu search, are implemented to solve the resulting scheduling problem. Bothalgorithms are tested in the context of a realistic air defence scenario.

94

Page 99: Conference Theme: Better Decisions for Africa

Why travelling salesmen should consider sleeping out

Angela L Rademeyer∗, Robert Bennetto & Alexander SloanOPSI Systems, South [email protected], [email protected]& [email protected]

Abstract

Many companies are confronted with the problem of creating fixed master routes fora period of more than a day for their geographically dispersed sales representatives,service technicians, home healthcare providers, etc. This involves the assignmentof the company’s customers to the service persons as well as service profiles. Aprofile indicates a valid combination of potential visit days for a customer as well asa proportion of distributable workload (time at customer) for each visit. Optimalroutes may include sleep-outs which are governed by rules indicating possible nightsspent away from home. The problem being modelled is an extension of a famouscombinatorial optimisation problem, the travelling salesman problem (TSP). Themodel includes the following features which differentiate it from work seen in theliterature:

(1) Modelling of more than one salesman (multiple TSP problem (mTSP)),

(2) Achieving periodic visits (periodic TSP (PTSP)),

(3) Allowing for the option to solve for the optimal number of salesmen and theirhome locations (location-allocation routing problem), and

(4) Suggesting sleep-out locations rather than choosing from an input list of hotels(TSPHS).

A genetic algorithm hybridised with problem specific heuristics (i.e. memetic algo-rithm) is applied to instances of this problem which cannot be solved using exactmethods in a reasonable amount of time. Real world data sets with different con-straints are used to test the algorithms on different problem sizes as well as classicalproblems from the literature where possible. The aim is to reduce the gap betweenexisting models for such nominated visit day (NVD) problems and the complex re-quirements of real applications.

95

Page 100: Conference Theme: Better Decisions for Africa

—List of Delegates* —

Adewumi, Aderemi O (University of KwaZulu-Natal, South Africa)Ajibola, Sunday A (University of KwaZulu-Natal, South Africa)Atobatele, Abisona S (Absat Engineering Ltd, Nigeria)Bashe, Mantombi (Eskom, South Africa)Bean, Wilna (Council for Scientific and Industrial Research, South Africa)Bester, Margarete J (XTranda, South Africa)Bezuidenhoudt, Cecile (University of Cape Town, South Africa)Blamah, Nachamada V (University of KwaZulu-Natal, South Africa)Bothma, Brahm (RTT Medical, South Africa)Brand, Andre (North-West University, South Africa)Breed, Gerbrand D (North-West University, South Africa)Buhrmann, Joke (OPSI Systems, South Africa)Campbell, Ian (University of the Witwatersrand, South Africa)Chinoda, Muriel (University of South Africa)Dannhauser, Louis F (Vodacom (Pty) Ltd, South Africa)De Villiers, Anton P (Stellenbosch University, South Africa)Dean, John F (Private Capacity, South Africa)Dempers, Clemens (Blue Stallion Technologies, South Africa)Du Toit, Jacques (Stellenbosch University, South Africa)Du Toit, Tiny (North-West University, South Africa)Durbach, Ian (University of Cape Town, South Africa)Einhorn, Mark D (Stellenbosch University, South Africa)Evans, David W (Development Bank of Southern Africa, South Africa)Fagan, Francois J (Stellenbosch University, South Africa)Fasondini, Mario (Sasol Mining, South Africa)Fatti, L Paul (University of the Witwatersrand, South Africa)Gevers, Wim R (Stellenbosch University, South Africa)Harmse, Martha FP (Sasol Synfuels, South Africa)Hearne, John W (Royal Melbourne Institute of Technology, Australia)Heyns, Andries M (Stellenbosch University, South Africa)Ittmann, Hans W (HWI Consulting, South Africa)Jankowitz, Mardi (University of South Africa)Jean-Robert, Kala K (Catholic University of Central Africa, Cameroon)Joubert, Jaco L (Sasol Shared Services, South Africa)Kapamara, Truword (Coventry University, United Kingdom)Kruger, Hennie (North-West University, South Africa)Kruger, Jan W (University of South Africa)Labuschagne, Pieter (North-West University, South Africa)Lane-Visser, Tanya E (Stellenbosch University, South Africa)Lindner, Berndt G (Stellenbosch University, South Africa)Lotter, Daniel P (Stellenbosch University, South Africa)

96

Page 101: Conference Theme: Better Decisions for Africa

Lutu, Patricia EN (University of Pretoria, South Africa)Magadla, Thandulwazi (University of Cape Town, South Africa)Masetshaba, Mantepu T (University of Limpopo, South Africa)Mauda, Rofhiwa (University of the Witwatersrand, South Africa)Moloi, Khehla D (University of Limpopo, South Africa)Mudhara, Brian (National University of Science & Technology, Zimbabwe)Munapo, Elias (University of South Africa)Mutasa, Farikayi K (National University of Science & Technology, Zimbabwe)Myeni, Senzosenkosi (Eskom, South Africa)Nare, Hausitoe (National University of Science & Technology, Zimbabwe)Ncube, Ozias (University of South Africa)Nel, Johanna H (Stellenbosch University, South Africa)Nkoane, Simon S (University of Limpopo, South Africa)Nzvengende, Tanyaradzwa S (National University of Science & Technology, Zimbabwe)Olusanya, Micheal O (University of KwaZulu-Natal, South Africa)Osei-Tutu, Angela NS (University of South Africa)Osiyemi, Olatunde O (University of Fort Hare, South Africa)Payne, Daniel F (Eskom, South Africa)Pelser, Winnie C (Arms Corporation of South Africa)Phillips, Colin A (OPSI Systems, South Africa)Pienaar, Wessel J (Stellenbosch University, South Africa)Pillay, Nelishia (University of KwaZulu-Natal, South Africa)Potgieter, Linke (Stellenbosch University, South Africa)Rajaratnam, Kanshukan (University of Cape Town, South Africa)Rapholo, Phineas (First National Bank, South Africa)Saul, Sandile (University of KwaZulu-Natal, South Africa)Schlunz, Evert B (South African Nuclear Energy Corporation)Sigauke, Caston (University of Limpopo, South Africa)Snyders, Frans J (Stellenbosch University, South Africa)Stewart, Theodor J (University of Cape Town, South Africa)Steyn, Hendrik J (Private Capacity, South Africa)Steyn, Tjaart (North-West University, South Africa)Terblanche, Stephanus E (North-West University, South Africa)Van der Merwe, Anette (Sasol Technology, South Africa)Van der Merwe, Annette (North-West University, South Africa)Van der Merwe, Michelle (Stellenbosch University, South Africa)Van der Westhuizen, Nica (Sasol Shared Services, South Africa)Van Vuuren, Jan H (Stellenbosch University, South Africa)Veldhuizen, Patrick R (Sasol Shared Services, South Africa)Vermaak, Ester J (Sasol Technology, South Africa)Viljoen, Nadia M (Council for Scientific and Industrial Research, South Africa)Wilgenbus, Erich F (North-West University, South Africa)Yoon, Moonyoung (Stellenbosch University, South Africa)

* Delegates who had already registered by September 1st, 2012.

97

Page 102: Conference Theme: Better Decisions for Africa

— Sponsors —

The Operations Research Society of South Africa gratefully acknowledges the sup-port and sponsorship provided by the following organisations:

• Blue Stallion Technologies, for its contribution towards partial sponsorship ofthe conference.

• Stellenbosch University, for sponsoring the travel costs of the keynote speaker,Professor John Hearne.

— Service Provision —

The conference organisers would like to thank the following organisations andindividuals for services provided:

• The printers, Minuteman Press, for printing this conference programme.

• The Meetings & Events Manager, Michelle Edgecomb, and her entire teamat the Aloe Ridge Hotel, for catering and making their facilities available toORSSA.

c© ORSSA (2012). Compiled and edited by Jan van Vuuren with considerableproofreading help by Anton de Villiers, Mark Einhorn and Danie Lotter.

98