MODELING AND ANALYSIS

1
MODELING AND ANALYSIS

• MN4118 Session 9
• Prof. Mark Nissen

2
Agenda

• Du Pont Rail Car Fleet
• Modeling Methods
• DSS in Action
• Examples
• Summary

3
Du Pont Rail Car Fleet

• Du Pont has rail car fleet
• Problem?
• Management intuition?
• DSS employed? Result?
• Best alternative?
• Military application of methods/tools?

4
Modeling

• Modeling art science?
• Which parts art?
• Example?
• Which parts science?
• Example?
• Model-design decisions
• Granularity, accuracy, size, dynamics, class,
  validation, others
• Auto-selection problem from Lab?

5
Quantitative Decision Methods

• Math programming - deterministic
• Single goal - stochastic
• Decision table/tree/influence diagram
• Regression/ANOVA
• Multiple goals - stochastic
• Multicriteria DM/MAU/AHP
• Multivariate statistics
• Numerical sim - experimental
• Neural nets - machine learning
• Others

6
Qualitative Decision Methods

• Convert qual factors to quant
• Utility (MAU)
• Preferences, tradeoffs equivalents
• Qualitative simulation
• Heuristic search
• Expert systems
• Genetic algorithms
• Symbolic machine learning
• Others

7
Decision Problems Modeling Approaches

• Decision model categories (T 5.1)
• Optimization few or many alternatives
• Simulation
• Heuristics
• Statistics
• Multi-attribute utility
• Match with problem characteristics?
• Certainty, structure, representation
  (qual/quant), accuracy, alternatives
• Matching exercise

8
DSS in Action

• Efes malt plant location
• Montreal snow removal planning
• PacBell network design
• Del Taco labor analysis
• Army Hospital staffing analysis

9
Influence Diagrams

• Model variables influence
• Visually understand model structure
• Decision variables
• Uncontrollable variables
• Results variables
• Model relationships
• Have we seen these before?
• Software to support
• Exercise route planning

10
Decision Trees

• Model variables influence
• Visually understand model structure
• Optimize few alternatives
• Software to support
• Exercise software contractor selection

11
Heuristics

• Rules of thumb
• Experience-based techniques
• Usually work but not guaranteed
• Applicable to satisficing problems
• Can be much faster than optimization
• Qualitative problems OK
• Address problems otherwise infeasible
• Automate via expert systems
• Example where to park

12
Simulation

• Virtual trial and error
• Experiment with model
• Controlled conditions
• Applicable to satisficing problems
• Explore ( viz) model dynamics
• Generally quantitative problems
• Address problems otherwise infeasible
• Automate via sim software
• Example ithink manpower model

13
Summary

• Modeling combines art and science
• Quantitative qualitative methods
• Select appropriate method for problem
• Exam questions?