doctor of science - operations research by weoptit · doctor of science - operations research ......

Post on 23-May-2020

3 Views

Category:

Documents

0 Downloads

Preview:

Click to see full reader

TRANSCRIPT

Optimization & ML by Weoptit

Tuomas LahtinenDep. CEO, Head of Development

Weoptit / Visma Consulting Finland

Doctor of Science - Operations Research

Expert in mathematical modeling, optimization and operational strategy

Weoptit in a nutshell

••

Optimization & Machine learning in Resource Management

Planning and allocation to ensure that…

the right amount of

the right kinds of resources, are in the right place,

at the right time

the right amount of

the right kinds of resources, are in the right place,

at the right time

The rostering puzzle - combinatorial optimization

Benefits of optimization in rostering

● Quality – Ergonomic rosters increase staff well being and productivity

● Efficiency – Reduced overtime compensations & use of flexible resources

● Automation – Less time spent, easier to react to changes

● Scenario analysis – “How many more we should hire to satisfy future demand”

Prediction problems in resource management

Demand for products and services● Sales success

● Orders per customer

● …

Supply of resources● Resignations

● Sickleaves

● …

Benefits of using ML & statistical techniques● Higher accuracy – more firm basis to prepare for the future

● Automation – less time spen

Examples

Rostering for care homes

Challenge

● Care business is very labor intensive -> thousands of rosters produced annually

for thousands of staff

● Difficult to create a roster that satisfies all the legal requirements, soft

constraints and is ergonomic and cost-effective

● Lot of time spent on planning by health care professionals

● The rosters could be better

Efficient high-quality rosters produced in less than an hour

Scheduling of tasks in car manufacturing

Challenge

● Planning of electrical tests is time consuming and difficult

● ~500 tests with complicated interdependencies

● Very complex optimization problem: Minimum time to complete all tests not

known

● Weoptit hired to challenge an internal team

Illustrative example

Product Manual Internal WeoptitProduct 1 800,2s 347,4s 241,8

Product 2 858,8s 259,5 241,8s

Product 3 854,3s 259,5 239,8s

Product 4 58,3s 52,5s 32,1s

Product 5 815,8s 310,9s 194,8s

Product 6 815,8s 310,9s 149,3s

Product 7 57,3s 51,9s 32,1s

Product 8 835,5s 278,6s 162,3s

Product 9 835,5s 278,6s 145,5s

Product 10 807s 278,5s 145,5s

Project win/loss prediction for a project based business

Challenge

● Business participates in tender competitions to win projects

● Projects rather homogenous, lots of data in CRM system

● Predicting win/loss to improve resource planning & purchasing

● Estimating the relationship between price and winrate supports pricing

75% accuracy in data with 50/50 wins/losses

Predictability to resourcing, purchasing of materialsOptimal pricing

Conclusion

Modern optimization & ML technologies create significant value in business development

Explore the full potential of your data with ML & optimization pilots

Together with Visma we can offer very strong full-scale solutions when a business case is found

top related