Decision Making
choose the optimum strategy from all the
Decision Making Situations
Perfect Information Maximize –
Minimize
Partial or Imperfect Information:
Decisions under Risk
Decisions under Uncertainty
Decisions under Risk
Based on criteria:
Expected value (of profit or loss)
Combined expected value and variance
Known aspiration level
Expected Value Criterion
Expected Value includes the probability to
gain profit + the probability to suffer loss
Expected Value and Variance
Expected Value + Variance determine the
Risk Aversion Factor (K)
Risk Aversion Factor indicates the
“importance” of an alternative
The higher value of K, the more important
Aspiration-Level Criterion
First alternative generally treats as the
“best” alternative
Most Likely Future Criterion
Simplification of probabilistic problem to
deterministic
Generalization of what happen in the
Probabilities for Under Risk
Prior Probabilities: the known probability Posterior Probabilities: modification of
Under Risk: Decision Tree
Nodes:
Square: decision point
Decision Tree: The Example
A company considers alternatives of
10-year plan (partitioned in 2-10-year and 8-10-year plans)
Stage 1: At the beginning of the 2-year
plan
Build large plant: – 5M
High Demand (prob. = 0.75) yields 1 M/year Low Demand (prob. = 0.25) yields 0.3 M/year
Build small plant: – 1M
Decision Tree
Stage 2: at the beginning of the 8-year
plan
Expand the small plant: – 4.2 M
High Demand (prob. = 0.75) yields 0.9 M/year Low Demand (prob. = 0.25) yields 0.2 M/year
Do not Expand the small plant:
2 3 1 M/y 0.3 M/y 0.75 0.25 Build larg e Plan t Bu ild sm all Pla nt H. demand L. demand H. de man d –5M – 1M 0.75
L. dem
and
0.25
0.2 M/y Stage 1: 2 years
Expa nd
1
4
Stage 2 : 8 years
(2)= (10 0.75 1) + (10 0.25 0.3) = 8.25
(1) (2)= 8.25 – 5 = 3.25 (build the large plant now)
(5)= (8 0.75 0.9) + (8 0.25 0.2) = 5.8
(6)= (8 0.75 0.25) + (8 0.25 0.2) =1.9
(4) (5),(6) = 5.8 + 1.9 – 4.2 = 3.5
(3) (4) = 3.5 + (2 0.75 0.25) + (10 0.25 0.20) = 4.375
Decision under Uncertainty
The Laplace Criterion optimistic
The Minimax (Maximin) Criterion less
optimistic
The Savage Criterion “less
conservative”
The Hurwicz Criterion ranging from
Decision under Uncertainty
Rows : possible action (ai)
Laplace Criterion
Based on the principle of insufficient reason
Unknown probabilities of the occurrence of
j = 1, 2, … n
j
Laplace Criterion
n j j aai
v
n
i 1)
,
(
1
max
n
1
Probability of
jLaplace Criterion: Example
15 19 22 30a
4 21 12 18 21a
3 23 8 7 8a
2 25 18 10 5a
1
4
3
2
1 Customer Category Supplies LevelProbability P{ =j} = ¼
j = 1, 2, 3, 4
• E{a1} = ¼ (5 + 10 + 18 + 25) = 14.5
• E{a2} = ¼ (8 + 7 + 8 + 23) = 11.5
• E{a3} = ¼ (21 + 18 + 12 + 21) = 18.0
Laplace Criterion: Example
15 19 22 30a
4 21 12 18 21a
3 23 8 7 8a
2 25 18 10 5a
1
4
3
2
1 Customer Category Supplies LevelMinimax (Maximin) Criterion
Making the best out of
Minimax Criterion: Example
Minimax Strategy 15 21 23 25 4v(ai, j)
30 19 22 30 a4 21 12 18 21 a3 23 8 7 8 a2 25 18 10 5 a1
max{v(ai, j)}
j
3 2
1
Maximin Criterion: Example
Maximin Strategy 15 21 23 25 4v(ai, j)
15 19 22 30 a4 12 12 18 21 a3 7 8 7 8 a2 5 18 10 5 a1
min{v(ai, j)}
j
3 2
1
Savage Minimax Regret Criterion
Construct new loss or profit matrix
v(ai, j) is replaced by r(ai, j) which is
defined by
max{v(ak, j)} – v(ai, j) if v is profit ak
r(ai, j)
v(ai, j) – min {v(ak, j)} if v is loss ak
Only the Minimax criterion can be applied
Savage Minimax Regret Criterion
15 21 23 25 4v(ai, j)
19 22 30 a4 12 18 21 a3 8 7 8 a2 18 10 5 a1 3 2 1
Savage Minimax Regret Criterion
15 19 22 30 a4 v(ai, j)15 21 23 25
4
min {v(ak, j)}
ak 5 7 8
12 18 21 a3 8 7 8 a2 13 10 5 a1 3 2 1
Savage Minimax Regret Criterion
0 6 8 10 4r(ai, j)
11 15 25 a4 4 11 16 a3 0 0 3 a2 10 3 0 a1 3 2 1
Savage Minimax Regret Criterion
0 6 8 10 4 25 16 8 10max r(ai, j)
j
r(ai, j)
11 15 25 a4 4 11 16 a3 0 0 3 a2 10 3 0 a1 3 2 1
v is loss
Hurwicz Criterion
Balancing between extreme pessimism
Hurwicz Criterion
v(ai, j) : profit or gain
max { max v(ai, j) + (1 – )min v(ai, j)
Hurwicz Criterion
v
(a
i,
j) : profit or gain
Most optimistic:
max max{v(ai, j)} ai jweigth =
Most pesimistic:
max min{v(ai, j)} ai jweigth = 1 –
where 0 1
Hurwicz Criterion
v
(a
i,
j) : cost
Most optimistic:
min min{v(ai, j)} ai jweigth =
Most pesimistic:
min max{v(ai, j)} ai jweigth = 1 –
where 0 1
Hurwicz Criterion
v(ai, j) : cost
min { min v(ai, j) + (1 – )max v(ai, j)
15 12 7 5
min v(ai, j)
j 15 21 23 25
4 30 19 22 30a
4 21 12 18 21a
3 23 8 7 8a
2 25 18 10 5a
1max v(ai, j)
j
3
2
1Minimize total cost of actions
+ 1 –
max v(ai, j)
j
0.5 15 + 0.5 30 =22.5
0.5 12 + 0.5 21 =16.5
0.5 7 + 0.5 23 =15 0.5 5 + 0.5 25 =15
min v(ai,
j) j
30 21 23 25
max v(ai, j)
j
15 12
7 5
min v(ai, j)
Reading Assignment #5
Game Theory
2 person zero-sum games
Reading Assignment #6
The End