4 smarter planet sig spatial spatial computing: recent trends

25

Upload: dinah-ferguson

Post on 13-Jan-2016

228 views

Category:

Documents


3 download

TRANSCRIPT

Page 1: 4 Smarter Planet SIG SPATIAL Spatial Computing: Recent Trends
Page 2: 4 Smarter Planet SIG SPATIAL Spatial Computing: Recent Trends
Page 3: 4 Smarter Planet SIG SPATIAL Spatial Computing: Recent Trends
Page 4: 4 Smarter Planet SIG SPATIAL Spatial Computing: Recent Trends

4

SmarterPlanet

SIG SPATIAL

Spatial Computing: Recent Trends

Page 5: 4 Smarter Planet SIG SPATIAL Spatial Computing: Recent Trends

Group MembersFacultyProfessor Shashi Shekhar

Current Ph.D. StudnetsPradeep MohanMike EvansDev OliverXun ZhouAbdussalam Bannur KwangSoo Yang Viswanath Gunturi   Zhe JiangJeff WolffChangqing Zhou

Others/VisitorsLydia ManikondaIvan Brugere

Group MembersFacultyProfessor Shashi Shekhar

Current Ph.D. StudnetsPradeep MohanMike EvansDev OliverXun ZhouAbdussalam Bannur KwangSoo Yang Viswanath Gunturi   Zhe JiangJeff WolffChangqing Zhou

Others/VisitorsLydia ManikondaIvan Brugere

Page 6: 4 Smarter Planet SIG SPATIAL Spatial Computing: Recent Trends

Ongoing Projects Overview

•ApplicationsTransportation, virtual environments, Earth science, epidemiology and cartography.•Spatial Data Mining

• Flow anomalies• Teleconnection• Cascade pattern discovery• K-Main-Route (KMR) summarization• Pattern of life• Abrupt change detection

•Spatial Database• Eco-Routing• Evacuation planning

Ongoing Projects Overview

•ApplicationsTransportation, virtual environments, Earth science, epidemiology and cartography.•Spatial Data Mining

• Flow anomalies• Teleconnection• Cascade pattern discovery• K-Main-Route (KMR) summarization• Pattern of life• Abrupt change detection

•Spatial Database• Eco-Routing• Evacuation planning

Page 7: 4 Smarter Planet SIG SPATIAL Spatial Computing: Recent Trends

Courses

Topics Application Domains Conceptual Data Models Logical Data Models Physical Data Models Spatial Networks Spatial Data Mining Others

Course Website http://www.spatial.cs.umn.edu/Courses/Fall11/8715

Topics Data Model Representation & access Architecture Others

Courses

Topics Application Domains Conceptual Data Models Logical Data Models Physical Data Models Spatial Networks Spatial Data Mining Others

Course Website http://www.spatial.cs.umn.edu/Courses/Fall11/8715

Topics Data Model Representation & access Architecture Others

CSCI 8715 – Spatial Databases and Applications

CSCI 5980 – GIS: a computational perspective

National Research Council

Page 8: 4 Smarter Planet SIG SPATIAL Spatial Computing: Recent Trends

Flow AnomaliesFlow Anomalies

ProblemProblem– Discover dominant time Discover dominant time

periods that exhibit periods that exhibit anomalous behavioranomalous behavior

Why is it hard?Why is it hard?– A single dominant time A single dominant time

period may have subsets period may have subsets that are not anomalousthat are not anomalous

No Dynamic No Dynamic ProgrammingProgramming

ContributionsContributions– A SWEET (Smart Window A SWEET (Smart Window

Enumeration and Enumeration and Evaluation of persistent-Evaluation of persistent-Thresholds) ApproachThresholds) Approach

88

http://www.esri.com/news/arcuser/0405/ss_crimestats2of2.html

Sensor 5

Sensor 1

Sensor 2

Sensor 4

Sensor 3

Ex. An Oil Spill

(Source: http://www.sfgate.com/cgi-bin/news/oilspill/busan)

(Source: Shingle Creek, MN Study Site)

J. M. Kang, S. Shekhar, C. Wennen, P. Novak, Discovering Flow Anomalies: A SWEET Approach, In the Eighth IEEE International Conference on Data Mining (ICDM '08), pp. 851-856, Pisa, Italy, December 15-19, 2008.

Page 9: 4 Smarter Planet SIG SPATIAL Spatial Computing: Recent Trends

ProblemProblem– Find remote Find remote

connectionsconnections ExampleExample

– El Niño in PacificEl Niño in Pacific Why is it hard?Why is it hard?

– Large spatial Large spatial datasetdataset

– Long time seriesLong time series

99

Dead Zone, Gulf of Mexico

Global Influence of El Nino during the Northern Hemisphere Winter (D: Dry, W: Warm, R: Rainfall)

TeleconnectionsTeleconnections

Page 10: 4 Smarter Planet SIG SPATIAL Spatial Computing: Recent Trends

Cascading spatio-temporal patterns (CSTPs)Cascading spatio-temporal patterns (CSTPs)Aggregate(T1,T2,T3)

Time T1

Assault(A)

Drunk Driving (C)Bar Closing(B)

Time T3>T2Time T2 > T1

a Input: Crime reports with location and time.

Output: Cascading spatio-temporal patterns

Courtsey: www.startribune.com Bar closing a generator for crime related CSTP!

Bar locations in Lincoln, NE

Why are CSTPs important ? Why is discovering CSTPs hard ? Trade off between computational efficiency and statistical interpretation. Pattern space exponential in number of event types.

Why are CSTPs Novel/better ? Current STDM literature ignores spatio-temporal semantics(e.g. partial order)

B A

C

CSTP: P1

Contributions Interest measure: Cascade participation index lower bound on conditional probability. Computational Structure

Compute measure efficiently Avoid unnecessary measure computations

Results:

{Bar Closing}

{Vandalism} {Assault}

CPI = 0.022; CPI-Downtown = 0.11

•Pradeep Mohan, Shashi Shekhar, James A. Shine, James P. Rogers. Cascading spatio-temporal pattern discovery: A summary of results. In Proc. of 10th SIAM International Data Mining (SDM) 2010, Columbus, OH, USA •Pradeep Mohan, Shashi Shekhar, James A. Shine, James P. Rogers. Cascading spatio-temporal pattern discovery. IEEE Transactions on Knowledge and data engineering(Accepted, In Press).

Page 11: 4 Smarter Planet SIG SPATIAL Spatial Computing: Recent Trends

Problem Statement:

The spatial network activity summarization (SNAS) problem: Given a spatial network and a collection of activities (e.g., crime reports, emergency requests), find a set of k paths to summarize the activities.

A K-Main Routes Approach to Spatial Network Activity A K-Main Routes Approach to Spatial Network Activity SummarizationSummarization

Importance:

SNAS is important for crime analysis and disaster response.

Challenge:Computational Complexity • Choose(N,2) paths, given N nodes• Exponential number of k subsets of paths

Contribution

The K-Main Routes (KMR) algorithm • Discovers k paths to summarize activities.• Generalizes K-means for network space but uses paths instead of ellipses to summarize activities. • Improves performance by using a network voronoi technique to assign activities to summary paths and a divide and conquer method to recompute summary paths.

Dev Oliver, Shashi Shekhar, James M. Kang, Renee Bousselaire, Abdussalam Bannur

Related Work:

Input K-Means Output KMR Output

KMR uses paths instead of ellipses in summarizing activities

Results•Proposed two new algorithms for improving the performance of KMR: Network Voronoi activity Assignment (NOVA) and Divide and conquer Summary PAth REcomputation (D-SPARE).•Validation via case studies, experiments and analytical evaluation to verify correctness in context of real workloads.•Successfully transferred software for direct evaluation by the National Geospatial-Intelligence Agency.

Input K-Means Output KMR Output

Page 12: 4 Smarter Planet SIG SPATIAL Spatial Computing: Recent Trends

Abrupt Change Interval DetectionAbrupt Change Interval Detection

Publication: Xun Zhou, Shashi Shekahr, Pradeep Mohan, Stefan Liess, Peter K. Snyder, Discovering Interesting Sub-paths in Spatiotemporal Datasets: A Summary of Results. In Proc. 19 th Intl’ Conf. Advances on Geographical Information Systems (ACM GIS 2011), Nov 2011, Chicago, IL, USA.

Given: A path A path SS in a Spatiotemporal Dataset in a Spatiotemporal Dataset

A unit-interval change abruptness threshold A unit-interval change abruptness threshold a

A sameness degree threshold A sameness degree threshold sd

Find:Dominant ST sub-intervals of Dominant ST sub-intervals of SS with with persistently abrupt changepersistently abrupt change

Objective:Reduce Computational Cost

Constraints:Constraints: Correctness & CompletenessCorrectness & Completeness

Vegetation cover in Africa, August 1-15, 1981.

Abrupt vegetation cover change in Africa, August 1-15, 1981.

Results:Results:Temporal intervals of Temporal intervals of abrupt rainfall change in abrupt rainfall change in Sahel, Africa.Sahel, Africa. Longitudinal spatial Longitudinal spatial abrupt change of abrupt change of vegetation cover in Africa.vegetation cover in Africa.

Page 13: 4 Smarter Planet SIG SPATIAL Spatial Computing: Recent Trends

Fuel Efficient Routing

Venkata M. V. Gunturi, Ernesto Nunes, KwangSoo Yang, and Shashi Shekhar. 2011. A critical-time-point approach to all-start-time lagrangian shortest paths: a summary of results. In SSTD'11, pp 74--91

INPUT: Road network; a source and destination; a time interval

OUTPUT: A path between source and destination for each start time

OBJECTIVE: The path should be fuel efficient.

Page 14: 4 Smarter Planet SIG SPATIAL Spatial Computing: Recent Trends

Evacuation PlanningEvacuation Planning

University of Minnesota 2006 Annual Report

(http://www.research.umn.edu/communications/publications/documents/OVPRAnnualRpt06.pdf)

Page 15: 4 Smarter Planet SIG SPATIAL Spatial Computing: Recent Trends

Evacuation Planning System in Cloud Environment

Given Transportation network with capacity constraints Initial number of people to be evacuated and their initial locations Evacuation destinations

Output Routes to be taken and scheduling of people on each route

Objective Minimize total time needed for evacuation Minimize computational overhead

Constraints Capacity constraints: evacuation plan meets capacity of the network Network data size is too large. (Data are stored into secondary storage) Utilize cloud environment for scalability

Problem Statement

Why Evacuation Planning?Hurricane Andrew Florida and Louisiana, 1992

( National Weather Services)

Hurricane Rita Gulf Coast, 2005 ( www.washingtonpost.com)

( National Weather Services)

( FEMA.gov)

System Architecture for Cloud Environment

Lack of effective evacuation plans Traffic congestions on all highways Great confusions and chaos

"We packed up Morgan City residents to evacuate in the a.m. on the day that Andrew hit coastal Louisiana, but in early afternoon the majority came back home. The traffic was so bad that they couldn't get through Lafayette." Mayor Tim Mott, Morgan City, Louisiana ( http://i49south.com/hurricane.htm )

Hurricane Rita evacuees from Houston clog I-45.

A Real Scenario (Monticello): Result Routes

Page 16: 4 Smarter Planet SIG SPATIAL Spatial Computing: Recent Trends

Spatial Computing in GovernmentSpatial Computing in Government

Page 17: 4 Smarter Planet SIG SPATIAL Spatial Computing: Recent Trends

Economy & Spatial Computing Economy & Spatial Computing

Page 18: 4 Smarter Planet SIG SPATIAL Spatial Computing: Recent Trends

Group Alumni

Academia:•Mete Celik (Erciyes Univ.)•Jin Soung Yoo (IU-Purdue Univ. Indy)•Hui Xiong (Rutgers Univ.)•Yan Huang (Univ. of North Texas)•Wei Li Wu (U. of Texas, Dallas)•Chang-Tien Lu (Virginia Polytechnic Univ)•Sanjay Chawla (Univ. of Sydney)•Du-Ren Liu (National Chiao Tung Univ.)•Andrew Yang (Univ. of Houston).Government Agency:•James Kang (USDOD)•Ranga Raju Vatsavai (USDOE-ORNL)Industry:•Betsy George (Oracle Spatial)•Qingsong Lu (Microsoft Research)•Sangho Kim (ESRI)•Baris Kazar (Oracle Spatial)•Pusheng Zhang (Microsoft Virtual Earth)•Xuan Liu (IBM TJ Watson)•Siva Ravada (Oracle)•Mark Coyle (Appirio)•Babak Hamidzadeh (Boeing Research)

Group Alumni

Academia:•Mete Celik (Erciyes Univ.)•Jin Soung Yoo (IU-Purdue Univ. Indy)•Hui Xiong (Rutgers Univ.)•Yan Huang (Univ. of North Texas)•Wei Li Wu (U. of Texas, Dallas)•Chang-Tien Lu (Virginia Polytechnic Univ)•Sanjay Chawla (Univ. of Sydney)•Du-Ren Liu (National Chiao Tung Univ.)•Andrew Yang (Univ. of Houston).Government Agency:•James Kang (USDOD)•Ranga Raju Vatsavai (USDOE-ORNL)Industry:•Betsy George (Oracle Spatial)•Qingsong Lu (Microsoft Research)•Sangho Kim (ESRI)•Baris Kazar (Oracle Spatial)•Pusheng Zhang (Microsoft Virtual Earth)•Xuan Liu (IBM TJ Watson)•Siva Ravada (Oracle)•Mark Coyle (Appirio)•Babak Hamidzadeh (Boeing Research)

Page 19: 4 Smarter Planet SIG SPATIAL Spatial Computing: Recent Trends

1919

Spatial/Spatio-temporal Data Mining: Representative Project

Nest locations Distance to open water

Vegetation durability Water depth

Location prediction: nesting sites Spatial outliers: sensor (#9) on I-35

Co-location Patterns Tele connections

(Ack: In collaboration w/V. Kumar, M. Steinbach, P. Zhang)

Page 20: 4 Smarter Planet SIG SPATIAL Spatial Computing: Recent Trends

2020

Spatial Databases: Representative Projects

only in old plan

Only in new plan

In both plans

Evacutation Route Planning

Parallelize Range Queries

Storing graphs in disk blocksShortest Paths

Page 21: 4 Smarter Planet SIG SPATIAL Spatial Computing: Recent Trends

Co-location PatternsCo-location Patterns

Yan Huang, Shashi Shekhar, and Hui Xiong, Discovering Co-location Patterns from Spatial Datasets: A General Approach, IEEE Transactions on Knowledge and Data Engineering (TKDE), 16(12), pp. 1472-1485, December 2004. (Earlier version appeared in SSTD ’01)

Given: A collection of different types of spatial event

Find:Co-located subsets of event types

Objective: Minimize computation time

Page 22: 4 Smarter Planet SIG SPATIAL Spatial Computing: Recent Trends

Spatial Outlier DetectionSpatial Outlier Detection

S. Shekhar, C.T. Lu, and P. Zhang. A unified approach to detecting spatial outliers. GeoInformatica, 7(2), 2003 (Earlier version appeared in SIGKDD ’01).

Given: A spatial graph G={V,E} A spatial graph G={V,E}

A neighbor relationship (K neighbors)A neighbor relationship (K neighbors)

An attribute function f : V -> RAn attribute function f : V -> R

An aggregation function : faggr :R k -An aggregation function : faggr :R k -> R> R

Confidence level threshold Confidence level threshold

Find:O = {vi | vi O = {vi | vi V, vi is a spatial outlier}V, vi is a spatial outlier}

Objective: Correctness: The attribute values of vCorrectness: The attribute values of vii

is extreme, compared with its is extreme, compared with its neighborsneighbors

Computational efficiencyComputational efficiency

Constraints:Constraints: Attribute value is normally distributed Attribute value is normally distributed

Computation cost dominated by I/O Computation cost dominated by I/O op.op.

Page 23: 4 Smarter Planet SIG SPATIAL Spatial Computing: Recent Trends

Nest locations Distance to open water

Vegetation durability Water depth

Location Prediction: Spatial Auto-regressionLocation Prediction: Spatial Auto-regression

S. Shekhar, P. Schrater, R. Vatsavai, W. Wu, and S. Chawla, Spatial Contextual Classification and Prediction Models for Mining Geospatial Data, In IEEE Transactions on Multimedia (special issue on Multimedia Dataabses) p174-188, 2002.

Given: Spatial Framework S={s1,…,sn}

Explanatory functions: fxi : S->R

A dependent class: fy : S->[0,1]

A family ζ of function mappings: R x…x R -> [0,1]

Find:

Classification model: f^

y Є ζ

Objective: Maximize classification accuracy

ConstraintsConstraints:: Spatial Autocorrelation exists

Page 24: 4 Smarter Planet SIG SPATIAL Spatial Computing: Recent Trends

Eco-Routing Eco-Routing

U.P.S. Embraces High-Tech Delivery Methods (July 12, 2007) By “The research at U.P.S. is paying off. ……..— saving roughly three million gallons

of fuel in good part by mapping routes that minimize left turns.”

• Minimize fuel consumption and GPG emission

– rather than proxies, e.g. distance, travel-time

– avoid congestion, idling at red-lights, turns and elevation changes, etc.

Do you idle at green light during traffic congestion?

Page 25: 4 Smarter Planet SIG SPATIAL Spatial Computing: Recent Trends

2525

Evacuation Planning: A Real Scenario, New Plan Routes

Source citiesDestination

Monticello Power Plant

Routes used only by old plan

Routes used only by result plan of capacity constrained routing

Routes used by both plans

Congestion is likely in old plan near evacuation destination due to capacity constraints. Our plan has richer routes near destination to reduce congestion and total evacuation time.

Twin Cities

Experiment Result

Total evacuation time:

- Existing Plan: 268 min.

- New Plan: 162 min.