chapter 6 multicriteria decision making · chapter 6 multicriteria decision making 2 goal...
TRANSCRIPT
1
Chapter 6Multicriteria Decision
Making
2
Goal Programming
Graphical Interpretation of Goal Programming
Computer Solution of Goal Programming Problems with QM for Windows and Excel
The Analytical Hierarchy Process
Scoring Models
Chapter Topics
2
3
Study of problems with several criteria, multiple criteria, instead of a single objective when making a decision.Three techniques discussed: goal programming, theanalytical hierarchy process and scoring models.Goal programming is a variation of linear programming considering more than one objective (goals) in the objective function.The analytical hierarchy process develops a score for each decision alternative based on comparisons of each under different criteria reflecting the decision makers preferences.Scoring models are based on a relatively simple weighted scoring technique.
Overview
4
Beaver Creek Pottery Company Example:
Maximize Z = $40x1 + 50x2
subject to:1x1 + 2x2 ≤ 40 hours of labor4x1 + 3x2 ≤ 120 pounds of clayx1, x2 ≥ 0
Where: x1 = number of bowls producedx2 = number of mugs produced
Goal Programming ExampleProblem Data (1 of 2)
3
5
Adding objectives (goals) in order of importance, thecompany:
Does not want to use fewer than 40 hours of labor per day.Would like to achieve a satisfactory profit level of $1,600 per day.Prefers not to keep more than 120 pounds of clay on hand each day.Would like to minimize the amount of overtime.
Goal Programming ExampleProblem Data (2 of 2)
6
All goal constraints are equalities that include deviational variables d- and d+.A positive deviational variable (d+) is the amount by which a goal level is exceeded.A negative deviation variable (d-) is the amount by which a goal level is underachieved.At least one or both deviational variables in a goal constraint must equal zero.The objective function in a goal programming model seeks to minimize the deviation from the respective goals in the order of the goal priorities.
Goal ProgrammingGoal Constraint Requirements
4
7
Labor goal:x1 + 2x2 + d1
- - d1+ = 40 (hours/day)
Profit goal:40x1 + 50 x2 + d2
- - d2+ = 1,600 ($/day)
Material goal:4x1 + 3x2 + d3
- - d3+ = 120 (lbs of clay/day)
Goal Programming Model FormulationGoal Constraints (1 of 3)
8
Labor goals constraint (priority 1 - less than 40 hours labor; priority 4 - minimum overtime):
Minimize P1d1-, P4d1
+
Add profit goal constraint (priority 2 - achieve profit of $1,600):
Minimize P1d1-, P2d2
-, P4d1+
Add material goal constraint (priority 3 - avoid keeping more than 120 pounds of clay on hand):
Minimize P1d1-, P2d2
-, P3d3+, P4d1
+
Goal Programming Model FormulationObjective Function (2 of 3)
5
9
Complete Goal Programming Model:Minimize P1d1
-, P2d2-, P3d3
+, P4d1+
subject to:x1 + 2x2 + d1
- - d1+ = 40 (labor)
40x1 + 50 x2 + d2 - - d2
+ = 1,600 (profit)4x1 + 3x2 + d3
- - d3 + = 120 (clay)
x1, x2, d1 -, d1
+, d2 -, d2
+, d3 -, d3
+ ≥ 0
Goal Programming Model FormulationComplete Model (3 of 3)
10
Changing fourth-priority goal “limits overtime to 10 hours”instead of minimizing overtime:
d1- + d4
- - d4+ = 10
minimize P1d1 -, P2d2
-, P3d3 +, P4d4
+
Addition of a fifth-priority goal- “important to achieve the goal for mugs”:
x1 + d5 - = 30 bowls
x2 + d6 - = 20 mugs
minimize P1d1 -, P2d2
-, P3d3 +, P4d4
+, 4P5d5 - + 5P5d6
-
Goal ProgrammingAlternative Forms of Goal Constraints (1 of 2)
6
11
Goal ProgrammingAlternative Forms of Goal Constraints (2 of 2)
Complete Model with Added New Goals:Minimize P1d1
-, P2d2-, P3d3
+, P4d4+, 4P5d5
- + 5P5d6-
subject to:x1 + 2x2 + d1
- - d1+ = 40
40x1 + 50x2 + d2- - d2
+ = 1,6004x1 + 3x2 + d3
- - d3+ = 120
d1+ + d4
- - d4+ = 10
x1 + d5- = 30
x2 + d6- = 20
x1, x2, d1-, d1
+, d2-, d2
+, d3-, d3
+, d4-, d4
+, d5-, d6
- ≥ 0
12
Minimize P1d1-, P2d2
-, P3d3+, P4d1
+
subject to: x1 + 2x2 + d1
- - d1+ = 40
40x1 + 50 x2 + d2 - - d2
+ = 1,6004x1 + 3x2 + d3
- - d3 + = 120
x1, x2, d1 -, d1
+, d2 -, d2
+, d3 -, d3
+ ≥ 0
Figure 9.1 Goal Constraints
Goal ProgrammingGraphical Interpretation (1 of 6)
7
13Figure 9.2 The First-Priority Goal: Minimize
Minimize P1d1-, P2d2
-, P3d3+, P4d1
+
subject to: x1 + 2x2 + d1
- - d1+ = 40
40x1 + 50 x2 + d2 - - d2
+ = 1,6004x1 + 3x2 + d3
- - d3 + = 120
x1, x2, d1 -, d1
+, d2 -, d2
+, d3 -, d3
+ ≥ 0
Goal ProgrammingGraphical Interpretation (2 of 6)
14Figure 9.3 The Second-Priority Goal: Minimize
Minimize P1d1-, P2d2
-, P3d3+, P4d1
+
subject to: x1 + 2x2 + d1
- - d1+ = 40
40x1 + 50 x2 + d2 - - d2
+ = 1,6004x1 + 3x2 + d3
- - d3 + = 120
x1, x2, d1 -, d1
+, d2 -, d2
+, d3 -, d3
+ ≥ 0
Goal ProgrammingGraphical Interpretation (3 of 6)
8
15Figure 9.4 The Third-Priority Goal: Minimize
Minimize P1d1-, P2d2
-, P3d3+, P4d1
+
subject to: x1 + 2x2 + d1
- - d1+ = 40
40x1 + 50 x2 + d2 - - d2
+ = 1,6004x1 + 3x2 + d3
- - d3 + = 120
x1, x2, d1 -, d1
+, d2 -, d2
+, d3 -, d3
+ ≥ 0
Goal ProgrammingGraphical Interpretation (4 of 6)
16Figure 9.5 The Fourth-Priority Goal: Minimize
Minimize P1d1-, P2d2
-, P3d3+, P4d1
+
subject to: x1 + 2x2 + d1
- - d1+ = 40
40x1 + 50 x2 + d2 - - d2
+ = 1,6004x1 + 3x2 + d3
- - d3 + = 120
x1, x2, d1 -, d1
+, d2 -, d2
+, d3 -, d3
+ ≥ 0
Goal ProgrammingGraphical Interpretation (5 of 6)
9
17
Goal programming solutions do not always achieve all goals and they are not “optimal”, they achieve the best or most satisfactory solution possible.Minimize P1d1
-, P2d2-, P3d3
+, P4d1+
subject to:x1 + 2x2 + d1
- - d1+ = 40
40x1 + 50 x2 + d2 - - d2
+ = 1,6004x1 + 3x2 + d3
- - d3 + = 120
x1, x2, d1 -, d1
+, d2 -, d2
+, d3 -, d3
+ ≥ 0
Solution: x1 = 15 bowlsx2 = 20 mugsd1
- = 15 hours
Goal ProgrammingGraphical Interpretation (6 of 6)
18Exhibit 9.4
Goal ProgrammingComputer Solution Using Excel (1 of 3)
10
19Exhibit 9.5
Goal ProgrammingComputer Solution Using Excel (2 of 3)
20Exhibit 9.6
Goal ProgrammingComputer Solution Using Excel (3 of 3)
11
21
Minimize P1d1-, P2d2
-, P3d3+, P4d4
+, 4P5d5- + 5P5d6
-
subject to:x1 + 2x2 + d1
- - d1+ = 40
40x1 + 50x2 + d2- - d2
+ = 1,6004x1 + 3x2 + d3
- - d3+ = 120
d1+ + d4
- - d4+ = 10
x1 + d5- = 30
x2 + d6- = 20
x1, x2, d1-, d1
+, d2-, d2
+, d3-, d3
+, d4-, d4
+, d5-, d6
- ≥ 0
Goal ProgrammingSolution for Altered Problem Using Excel (1 of 6)
22Exhibit 9.7
Goal ProgrammingSolution for Altered Problem Using Excel (2 of 6)
12
23Exhibit 9.8
Goal ProgrammingSolution for Altered Problem Using Excel (3 of 6)
24Exhibit 9.9
Goal ProgrammingSolution for Altered Problem Using Excel (4 of 6)
13
25Exhibit 9.10
Goal ProgrammingSolution for Altered Problem Using Excel (5 of 6)
26Exhibit 9.11
Goal ProgrammingSolution for Altered Problem Using Excel (6 of 6)
14
27
AHP is a method for ranking several decision alternatives and selecting the best one when the decision maker has multiple objectives, or criteria, on which to base the decision.The decision maker makes a decision based on how the alternatives compare according to several criteria.The decision maker will select the alternative that best meets his or her decision criteria.AHP is a process for developing a numerical score to rank each decision alternative based on how well the alternative meets the decision maker’s criteria.
Analytical Hierarchy ProcessOverview
28
Southcorp Development Company shopping mall site selection.Three potential sites:
AtlantaBirmingham Charlotte.
Criteria for site comparisons:Customer market base.Income levelInfrastructure
Analytical Hierarchy ProcessExample Problem Statement
15
29
Top of the hierarchy: the objective (select the best site).
Second level: how the four criteria contribute to the objective.
Third level: how each of the three alternatives contributes to each of the four criteria.
Analytical Hierarchy ProcessHierarchy Structure
30
Mathematically determine preferences for sites with respect to each criterion.
Mathematically determine preferences for criteria (rank order of importance).
Combine these two sets of preferences to mathematically derive a composite score for each site.
Select the site with the highest score.
Analytical Hierarchy ProcessGeneral Mathematical Process
16
31
In a pairwise comparison, two alternatives are compared according to a criterion and one is preferred.
A preference scale assigns numerical values to different levels of performance.
Analytical Hierarchy ProcessPairwise Comparisons (1 of 2)
32Table 9.1 Preference Scale for Pairwise Comparisons
Analytical Hierarchy ProcessPairwise Comparisons (2 of 2)
17
33
Income Level Infrastructure Transportation
A
B
C ⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢
⎣
⎡
1931/911/61/361
⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢
⎣
⎡
11/7171311/31
⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢
⎣
⎡
11/42413
1/21/31
Customer Market Site A B C A B C
1 1/3 1/2
3 1 5
2 1/5 1
Analytical Hierarchy ProcessPairwise Comparison Matrix
A pairwise comparison matrix summarizes the pairwise comparisons for a criteria.
34
Customer Market Site A B C
A B C
1 1/3
1/2 11/6
3 1 5 9
2 1/5 1
16/5
Customer Market Site A B C
A B C
6/11 2/11 3/11
3/9 1/9 5/9
5/8 1/16 5/16
Analytical Hierarchy ProcessDeveloping Preferences Within Criteria (1 of 3)
In synthesization, decision alternatives are prioritized with each criterion and then normalized:
18
35
Table 9.2 The Normalized Matrix with Row Averages
Analytical Hierarchy ProcessDeveloping Preferences Within Criteria (2 of 3)The row average values represent the preference vector
36Table 9.3 Criteria Preference Matrix
Analytical Hierarchy ProcessDeveloping Preferences Within Criteria (3 of 3)Preference vectors for other criteria are computed similarly,resulting in the preference matrix
19
37
Criteria Market Income Infrastructure Transportation Market Income Infrastructure Transportation
1 5
1/3 1/4
1/5 1
1/9 1/7
3 9 1
1/2
4 7 2 1
Table 9.4 Normalized Matrix for Criteria with Row Averages
Analytical Hierarchy ProcessRanking the Criteria (1 of 2)Pairwise Comparison Matrix:
38
⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
0.06120.08600.65350.1993
Analytical Hierarchy ProcessRanking the Criteria (2 of 2)
Preference Vector for Criteria:MarketIncomeInfrastructureTransportation
20
39
Overall Score:Site A score = .1993(.5012) + .6535(.2819) + .0860(.1790) + .0612(.1561) = .3091Site B score = .1993(.1185) + .6535(.0598) + .0860(.6850) + .0612(.6196) = .1595Site C score = .1993(.3803) + .6535(.6583) + .0860(.1360) + .0612(.2243) = .5314
Overall Ranking:
Site Score Charlotte Atlanta
Birmingham
0.5314 0.3091 0.1595 1.0000
Analytical Hierarchy ProcessDeveloping an Overall Ranking
40
Analytical Hierarchy ProcessSummary of Mathematical Steps
Develop a pairwise comparison matrix for each decision alternative for each criteria.Synthesization
Sum the values of each column of the pairwise comparison matrices.Divide each value in each column by the corresponding column sum.Average the values in each row of the normalized matrices.Combine the vectors of preferences for each criterion.
Develop a pairwise comparison matrix for the criteria.Compute the normalized matrix.Develop the preference vector.Compute an overall score for each decision alternativeRank the decision alternatives.
21
41
Analytical Hierarchy Process: Consistency (1 of 3)Consistency Index (CI): Check for consistency and validity of
multiple pairwise comparisonsExample: Southcorp’s consistency in the pairwise comparisons of the 4
site selection criteria
Step 1: Multiply the pairwise comparison matrix of the 4 criteria by its preference vector
Market Income Infrastruc. Transp. CriteriaMarket 1 1/5 3 4 0.1993Income 5 1 9 7 X 0.6535
Infrastructure 1/3 1/9 1 2 0.0860Transportation 1/4 1/7 1/2 1 0.0612
(1)(.1993)+(1/5)(.6535)+(3)(.0860)+(4)(.0612) = 0.8328(5)(.1993)+(1)(.6535)+(9)(.0860)+(7)(.0612) = 2.8524
(1/3)(.1993)+(1/9)(.6535)+(1)(.0860)+(2)(.0612) = 0.3474(1/4)(.1993)+(1/7)(.6535)+(1/2)(.0860)+(1)(.0612) = 0.2473
42
Analytical Hierarchy Process: Consistency (2 of 3)
Step 2: Divide each value by the corresponding weight from the preference vector and compute the average
0.8328/0.1993 = 4.17862.8524/0.6535 = 4.36480.3474/0.0860 = 4.04010.2473/0.0612 = 4.0422
16.257Average = 16.257/4
= 4.1564
Step 3: Calculate the Consistency Index (CI)CI = (Average – n)/(n-1), where n is no. of items compared
CI = (4.1564-4)/(4-1) = 0.0521
(CI = 0 indicates perfect consistency)
22
43
Analytical Hierarchy Process: Consistency (3 of 3)
Step 4: Compute the Ratio CI/RIwhere RI is a random index value obtained from Table 9.5
Table 9.5 Random Index Values for n Items Being Compared
CI/RI = 0.0521/0.90 = 0.0580 Note: Degree of consistency is satisfactory if CI/RI < 0.10
44Exhibit 9.12
Analytical Hierarchy ProcessExcel Spreadsheets (1 of 4)
23
45Exhibit 9.13
Analytical Hierarchy ProcessExcel Spreadsheets (2 of 4)
46Exhibit 9.14
Analytical Hierarchy ProcessExcel Spreadsheets (3 of 4)
24
47Exhibit 9.15
Analytical Hierarchy ProcessExcel Spreadsheets (4 of 4)
48
Each decision alternative graded in terms of how well it satisfies the criterion according to following formula:
Si = Σgijwj
where:wj = a weight between 0 and 1.00 assigned to criterion j; 1.00 important, 0 unimportant; sum of total weights equals one.gij = a grade between 0 and 100 indicating how well alternative i satisfies criteria j; 100 indicates high satisfaction, 0 low satisfaction.
Scoring ModelOverview
25
49
Mall selection with four alternatives and five criteria:
S1 = (.30)(40) + (.25)(75) + (.25)(60) + (.10)(90) + (.10)(80) = 62.75S2 = (.30)(60) + (.25)(80) + (.25)(90) + (.10)(100) + (.10)(30) = 73.50S3 = (.30)(90) + (.25)(65) + (.25)(79) + (.10)(80) + (.10)(50) = 76.00S4 = (.30)(60) + (.25)(90) + (.25)(85) + (.10)(90) + (.10)(70) = 77.75
Mall 4 preferred because of highest score, followed by malls 3, 2, 1
Grades for Alternative (0 to 100) Decision Criteria
Weight (0 to 1.00)
Mall 1
Mall 2
Mall 3
Mall 4
School proximity Median income Vehicular traffic Mall quality, size Other shopping
0.30 0.25 0.25 0.10 0.10
40 75 60 90 80
60 80 90 100 30
90 65 79 80 50
60 90 85 90 70
Scoring ModelExample Problem
50Exhibit 9.16
Scoring ModelExcel Solution
26
51
Goal Programming Example ProblemProblem Statement
Public relations firm survey interviewer staffing requirements determination.One person can conduct 80 telephone interviews or 40 personal interviews per day.$50/ day for telephone interviewer; $70 for personal interviewer.Goals (in priority order):
At least 3,000 total interviews.Interviewer conducts only one type of interview each day. Maintain daily budget of $2,500.At least 1,000 interviews should be by telephone.
Formulate a goal programming model to determine number of interviewers to hire in order to satisfy the goals, and then solve the problem.
52
Purchasing decision, three model alternatives, three decision criteria.Pairwise comparison matrices:
Prioritized decision criteria:
Price Bike X Y Z
X Y Z
1 1/3 1/6
3 1
1/2
6 2 1
Gear Action Bike X Y Z
X Y Z
1 3 7
1/3 1 4
1/7 1/4 1
Weight/Durability Bike X Y Z
X Y Z
1 1/3 1
3 1 2
1 1/2 1
Criteria Price Gears Weight Price Gears Weight
1 1/3 1/5
3 1
1/2
5 2 1
Analytical Hierarchy Process Example ProblemProblem Statement
27
53
Step 1: Develop normalized matrices and preference vectors for all the pairwise comparison matrices for criteria.
Price Bike X Y Z Row Averages
X Y Z
0.6667 0.2222 0.1111
0.6667 0.2222 0.1111
0.6667 0.2222 0.1111
0.6667 0.2222 0.1111 1.0000
Gear Action Bike X Y Z Row Averages
X Y Z
0.0909 0.2727 0.6364
0.0625 0.1875 0.7500
0.1026 0.1795 0.7179
0.0853 0.2132 0.7014 1.0000
Analytical Hierarchy Process Example ProblemProblem Solution (1 of 4)
54
Weight/Durability Bike X Y Z Row Averages
X Y Z
0.4286 0.1429 0.4286
0.5000 0.1667 0.3333
0.4000 0.2000 0.4000
0.4429 0.1698 0.3873 1.0000
Criteria Bike Price Gears Weight
X Y Z
0.6667 0.2222 0.1111
0.0853 0.2132 0.7014
0.4429 0.1698 0.3873
Step 1 continued: Develop normalized matrices and preference vectors for all the pairwise comparison matrices for criteria.
Analytical Hierarchy Process Example ProblemProblem Solution (2 of 4)
28
55
Step 2: Rank the criteria.
Price GearsWeight ⎥
⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
0.12220.22990.6479
Criteria Price Gears Weight Row Averages Price Gears Weight
0.6522 0.2174 0.1304
0.6667 0.2222 0.1111
0.6250 0.2500 0.1250
0.6479 0.2299 0.1222 1.0000
Analytical Hierarchy Process Example ProblemProblem Solution (3 of 4)
56
Step 3: Develop an overall ranking.
Bike XBike YBike Z
Bike X score = .6667(.6479) + .0853(.2299) + .4429(.1222) = .5057Bike Y score = .2222(.6479) + .2132(.2299) + .1698(.1222) = .2138Bike Z score = .1111(.6479) + .7014(.2299) + .3873(.1222) = .2806
Overall ranking of bikes: X first followed by Z and Y (sum of scores equal 1.0000).
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
•1222.02299.06479.0
3837.07014.01111.01698.02132.02222.04429.00853.06667.0
Analytical Hierarchy Process Example ProblemProblem Solution (4 of 4)