Principles of Modeling

Download Report

Transcript Principles of Modeling

PRINCIPLES OF
MODELING
Principle 1: Model Simple, Think complicated
Model
Rigorous Argumentation
Critical Thinking
Analysis
Principle 1: Model Simple, Think complicated
Routine use
Human interaction
Decision
automation
Routine
decision
support
System
investigation &
improvement
Models that
provide insight
Principle 2: be parsimonious – start small & add
KISS [keep it simple, stupid]
Ball Example
Principle 3: divide, conquer, avoid mega models
“Beware of general purpose, grandiose models that try to in
incorporate practically everything. Such models are difficult to
validate, to interpret, to calibrate statistically and, most
importantly, to explain. You may be better off not with one big
model but with a set of simpler models”.
Raiffe 1982
Example: manufacturer of packaging items that wishes to
provide a better service to geographically scattered customers.
Demand model
Production model
Customer priorities
Truck capacities
Principle 4: use metaphors, analogies and
similarities
Look for analogies based on previous experience – on well
developed logical structures that worked before.
Electricity Example
• Water flowing through pumps, valves and reservoirs - batteries
• People moving through crowd - resistance
Principle 5: do not fall in love with data
The model should drive the data, not vice-versa
Data mining and data grubbing
Beware of data provided in a plate
Data are just a sample
Avoid using the same data to build and test a model
Principle 6: Model building may feel like
muddling through
“develop their models, not in one burst, but over an
extended period of time …”
“…guided by analogies, drawings, doodling …”
60% time – model structure
30% time – problem context & model assessment
10% time – model realization
Willemain 1995
Summary
Principle 1: Model Simple, Think complicated
Principle 2: be parsimonious – start small & add
Principle 3: divide, conquer, avoid mega models
Principle 4: use metaphors, analogies and similarities
Principle 5: do not fall in love with data
Principle 6: Model building may feel like muddling through