Decision Analysis, Part 1

• Introduction to Decision Analysis

– Concepts: modelling, evaluation, analysis
– Decision Problem-Solving: Stages
– Relation of DA to some other Disciplines
Decision Analysis • Decision-Making under Uncertainty
Part 1 – Decision-Making under Strict Uncertainty
• Decision Table
• Various Decision Criteria
– Decision-Making under Risk
• Expected Value
• Sensitivity Analysis

Marko Bohanec Marko Bohanec

Decision Analysis Decision-Making Problem

Decision Analysis: Applied Decision Theory
options
Provides a framework for analyzing decision problems by goals (objectives)
(alternatives)
• structuring and breaking them down into more
manageable parts,
• explicitly considering the:
– possible alternatives, • FIND the option that best satisfies the
– available information goals
– uncertainties involved, and
– relevant preferences
• RANK options according to the goals
• combining these to arrive at optimal (or "good") decisions
• ANALYSE, JUSTIFY, EXPLAIN, …,
the decision
Marko Bohanec Marko Bohanec

Evaluation Models Types of Models in Decision Analysis

Decision Trees Multi-Attribute
options Utility Models
Succeed
Invest
EVALUATION Fail
Investment
Do not invest

EVALUATION
Costs Risks Results
MODEL Influence Diagrams

Invest? Success?

Analytic Hierarchy Process

Return
ANALYSIS

Marko Bohanec Marko Bohanec

•1
 Multi-Attribute Models Decision-Making Process
cars
buying

maint
PRICE

safety
CAR
doors
TECH

pers COMF

lug
problem decomposition

Marko Bohanec Source: Decision Analysis – A Tool to Deal with Uncertainty, http://www.dmreview.com/article_sub.cfm?articleId=6935 Marko Bohanec

Decision-Making Process The Decision Analysis Process

• Fact Finding Identify decision situation
INTELLIGENCE • Problem/Opportunity Sensing and understand objectives
• Analysis/Exploration
Identify alternatives

• Formulation of Solutions Decompose and model

• problem structure
DESIGN • Generation of Alternatives
• uncertainty
• Modelling/Simulation • preferences

Sensitivity Analyses
• Alternative Selection
CHOICE • Goal Maximization Choose best alternative

• Decision Making
Implement Decision
• Implementation
Marko Bohanec Marko Bohanec

Decision Analysis: Related Disciplines Decision Support and Decision Modeling

DSS, GDSS

• Group Decision Process Decision Support

Decision Trees Multi-Attribute
⇐ OR/MS Utility Models ES, ML
Modelling Decision Analysis
Succeed
Invest
Fail
Investment

• Multi-Criteria Do not invest

• Qualitative
Optimisation Influence Diagrams
Costs Risks Results
Multi-Attribute
• Risk Analysis and Invest? Success? Models
Simulation Analytic Hierarchy Process
• Bayesian Networks Return

• Markov Modelling
• Decision Tables
• Outranking
• Multi-Criteria Partial Ordering
• Mathematical/Algebraic/Statistical Modelling
• Accounting / Financial Modelling
Modelling


Marko Bohanec Marko Bohanec

•2
 Decision-Making Problem
Suppose that one must choose between
several uncertain alternatives.

Given:
Decision-Making under Uncertainty • Alternatives;
• The consequences of choosing each alternative,
described with a single number,
e.g. profit / loss in € or aggregated value.

Task: Which alternative to choose?

Marko Bohanec Marko Bohanec

Decision Table Working Example

Decision-Making under Strict Uncertainty A manufacturing company, faced with a possible increase
State of the world Value of alternatives 1 … m in demand for its product, considers the following:
(Event)
θ a1 ... am
Alternatives:
θ1 y11 ... y1m
1. status quo: no change
: : : 2. extend: extending their production line buying a new machine
θn yn1 ... ynm 3. build: building a new production hall with new equipment
4. cooperate: finding additional business parters for production
Decision-Making under Risk
State of the world Probability that θ will happen Value of alternatives 1 … m Uncertainty involved:
(Event)
θ P(θ) a1 ... am Market reaction: after the decision, the sales can increase or decrease.
θ1 p(θ1) y11 ... y1m
Consequences:
: : : : Expected profit, shown in decision table on the next slide
θn p(θn) yn1 ... ynm
Marko Bohanec Marko Bohanec

Working Example
Decision table

alternative
status quo extend build cooperate
decreased
sales
28 24 16 30 Decision-Making under Strict Uncertainty
states

increased
30 42 44 34
sales

Marko Bohanec Marko Bohanec

•3
 Decision Criteria Dominance
• Dominance • Choose the alternative with best consequences in all
• Pessimistic (Maximin, Wald’s) states of the world.
• Optimistic (Maximax) • Such alternative is seldom found.
• Hurwiz’s
• Laplace’s alternative
• Minimax Regret status quo extend build cooperate
decreased
28 24 16 30
sales

states
increased
30 42 44 34
sales

No dominant alternatives in this case

Marko Bohanec Marko Bohanec

Pessimistic Criterion (Wald’s, Maximin) Optimist’s Criterion (Maximax)

• Each alternative is represented by its worst possible • Each alternative is represented by its best possible
consequence. consequence.
• According to these, the alternative with the best worst • The alternative for which this best consequence is best
case is chosen. is chosen.
alternative alternative
status quo extend build cooperate status quo extend build cooperate
decreased decreased
28 24 16 30 28 24 16 30
sales sales
states

states

increased increased
30 42 44 34 30 42 44 34
sales sales
Pessimist 28 24 16 30 Optimist 30 42 44 34

Marko Bohanec Marko Bohanec

Hurwiz’s Criterion Hurwiz’s Criterion

• Introduce a parameter d ∈ [0,1]. 50

• Combine Optimistic and Pessimistic criteria so that

Evaluation of Alternatives (Hurwiz)

40

uh = duo + (1 − d )u p
30 status quo
extend
alternative build
status quo extend build cooperate 20
cooperate
decreased
28 24 16 30
sales
states

10
increased
30 42 44 34
sales
0
Pessimist 28 24 16 30 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1
Optimist 30 42 44 34 d

Hurwiz (d=0,3) 28,6 29,4 24,4 31,2

Marko Bohanec Marko Bohanec

•4
 Laplace’s Criterion Minimax Regret
• Consider all states (events) equally likely, The regret rij for the alternative aj in state θi is equal to
• thus, consider the average of outcomes for each the difference between the best alternative in given
alternative. state θi and aj: r = max m
{y } − y
ij ik ij
k =1

Choose the alternative having the least maximum regret.

alternative
status quo extend build cooperate alternative
decreased status quo extend build cooperate
28 24 16 30
sales
states

decreased
30–28=2 30–24=6 30–16=14 30–30=0
increased sales

States
30 42 44 34
sales increased
44–30=14 44–42=2 44–44=0 44–34=10
Laplace 29 33 30 32 sales
Regret 14 6 14 10

Marko Bohanec Marko Bohanec

Summary Questions
alternative
status quo extend build cooperate • If you were the manager, which alternative would you
decreased take? Why?
28 24 16 30
sales
states

increased
sales
30 42 44 34 • Is this really the best alternative? Why? Under which
Pessimist 28 24 16 30 circumstances it is best?
Optimist 30 42 44 34
Hurwiz (d=0,3) 28,6 29,4 24,4 31,2 • What can you say about the status quo alternative?
Laplace 29 33 30 32 According to the analysis, when should be it taken, or
Regret 14 6 14 10 should it be taken at all?

Marko Bohanec Marko Bohanec

Questions

Assess the presented decision criteria:

• Describe the prevalent characteristics of each criterion
• What do you think about the criteria:
– Are they comprehensible?
– Are they realistic? Decision-Making under Risk
– Are they useful for practice?
– Which is your favourite criterion?
• Is there a single “best” criterion? Which and why?

Marko Bohanec Marko Bohanec

•5
 Working Example Decision Criteria
Now we know (or estimate) the probablity of states • Mode: Select the most probable state
• Expected Value (EV), Expected Monetary Value (EMV)
alternatives
states probability status quo extend build cooperate
decreased
25% 28 24 16 30
sales
increased sales 75% 30 42 44 34

Marko Bohanec Marko Bohanec

Expected (Monetary) Value Sensitivity Analysis

50
n
Maximise the expected value: EVi = p( j )y ji
Expected value of alternatives

40
j =1

status quo
alternatives 30
extend
states probability status quo extend build cooperate
build
decreased 20
25% 28 24 16 30 cooperate
sales
increased sales 75% 30 42 44 34 10
expected
probability of
0,25×28+ 0,25×24+ 0,25×16+ 0,25×30+ stable area
increased sales
Expected value 0,75×30= 0,75×42= 0,75×44= 0,75×34= (75%)
0
29,5 37,5 37 33 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1
probability of increased sales

Marko Bohanec Marko Bohanec

Exercise 1 Exercise 2
P(θ) a1 a2 a3 Help the farmer who is deciding which crop to plant in
θ1 2/9 8 4 20 the face of uncertain weather and resulting crop
θ2 3/9 7 15 10 yield:
θ3 4/9 6 5 0
Weather
probability .55 .15 30
Given this decision table:
Normal Drought Rainy
• Determine which alternative is best according to all the
criteria (Dominance, Pessimistic, Optimistic, Hurwiz (d=0.7),
Plant soybeans $ 10 5 12
Laplace, Regret, Mode, Expected Value).
Plant corn 7 8 13
• Draw a chart evaluating the Hurwiz’s criterion for d ∈ [0,1].
• Do sensitivity analysis. profit per acre

Marko Bohanec Marko Bohanec

•6
 Exercise 3 Exercise 4

1. Define a decision problem of your own, Using some decision table,

2. represent it in a decision table,
3. and repeat the steps of Exercise 1
implement in spreadsheet software (such as MS Excel):
• evaluation of alternatives using all the criteria,
• drawing the chart associated with Hurwiz’s criterion
• drawing the sensitivity analysis chart

Compare the two charts.

Marko Bohanec Marko Bohanec

•7