Sie sind auf Seite 1von 15

Fuzzy Analytic Hierarchy Process (F-AHP)(Chang, 1996)

• MCDM Technique embedding fuzzy theory to AHP

• fuzzifies hierarchical analysis and allows fuzzy


numbers for pairwise comparisons to find fuzzy
preference weights.

Nisha S. Simon
Need ?
 Standard Questionnaire
 Relative Comparison among elements
 Attempt complex hierarchical models
 Consistency Measure
 Related to
› Structural Equation Modeling – No Relative
Comparison
› Conjoint Analysis – Higher and complex
combinations are difficult
Process

 Input – Subjective Opinion, Actual Measurement (Price, Weights);


provide judgments about the relative importance of each criterion and then
specify a preference for each decision alternative on each criterion
 Output – Prioritized ranking indicating the overall preference
for each of the decision alternatives, Consistency Index
 Process –
 Build hierarchical model (Criteria, Sub Criteria, Alternatives)
 Build pair wise comparison matrix for each expert
 Convert crisp numbers of each matrix to triangular fuzzy numbers and
their reciprocals (linguistic scale)
 Consolidate matrices using geometric mean method
 Calculate Fuzzy synthetic extent for each factor
 Find the degree of possibility values
 Obtain priority weights for each factor
 Convert them into normalized priority weights
 Check consistency ratio
Triangular Fuzzy Conversion Scale/ Linguistic Scale (Fuzzy Fundamental Scale 1 by Yazdani - Chamzini
and Yakhchali 2012; Soroor et al. 2012)

Linguistic Scale Crisp Number Triangular Fuzzy Reciprocal TFN


Number (TFN)
Equally Important 1 (1,1,1) (1,1,1)
Intermediate Value 2 (1,2,3) (1/3,1/2,1)
Weakly Important 3 (2,3,4) (1/4,1/3,1/2)
Intermediate Value 4 (3,4,5) (1/5,1/4,1/3)
Fairly Important 5 (4,5,6) (1/6,1/5,1/4)
Intermediate Value 6 (5,6,7) (1/7,1/6,1/5)
Strongly Important 7 (6,7,8) (1/8,1/7,1/6)
Intermediate Value 8 (7,8,9) (1/9,1/8,1/7)
Absolutely Important 9 (8,9,10) (1/10,1/9,1/8)
 Pair wise comparison matrix derived from
Questionnaire
Expert 1 Expert 2
Criteria C1 C2 C3 Criteria C1 C2 C3
C1 1 3 1 C1 1 3 1
C2 1 1/3 C2 1 1
C3 1 C3 1

Expert 3 Expert 3
Criteria C1 C2 C3 Criteria C1 C2 C3
C1 1 1/3 1/3 C1 1 1 1
C2 1 3 C2 1 1
C3 1 C3 1
 Pair wise comparison matrix converted to TFN
matrix(using fuzzy scale)
Expert 1
Criteria C1 C2 C3
C1 1,1,1 2,3,4 1,1,1
C2 1/4,1/3,1/2 1,1,1 1/4,1/3,1/2
C3 1,1,1 2,3,4 1,1,1
Expert 4
Expert 2 Criteria C1 C2 C3
Criteria C1 C2 C3 C1 1,1,1 1,1,1 1,1,1
C1 1,1,1 2,3,4 1,1,1 C2 1,1,1 1,1,1 1,1,1
C2 1/4,1/3,1/2 1,1,1 1,1,1 C3 1,1,1 1,1,1 1,1,1
C3 1,1,1 1,1,1 1,1,1

Expert 3
Criteria C1 C2 C3
C1 1,1,1 1/4,1/3,1/2 1/4,1/3,1/2
C2 2,3,4 1,1,1 2,3,4
C3 2,3,4 1/4,1/3,1/2 1,1,1
 Consolidated Pair wise Comparison Matrix using
Geometric Mean

Above table is converted using geometric mean method of aggregation


{ex: Value of C1C2 is derived as (GM of (2, 2,1/4,1), GM of (3,3,1/3,1),GM of(4,4,1/2,1)]
Criteria C1 C2 C3

C1 1,1,1 1,1.32,1.68 .71,.76,.84

C2 .59,.76,1 1,1,1 .84,1,1.19

C3 1.19,1.32,1.41 .84,1,1.19 1,1,1


 Fuzzy synthetic extent for each criterion (Si)
Criteria C1 C2 C3 Sum*

C1 1,1,1 1,1.32,1.68 .71,.76,.84 2.71,3.08,3.52

C2 .59,.76,1 1,1,1 .84,1,1.19 2.43,2.76,3.19

C3 1.19,1.32,1.41 .84,1,1.19 1,1,1 3.03,3.32,3.60

Sum 8.17,9.16,10.31

Inverse of sum and write in reverse order .10,.11,.12

Sum Inverse Synthetic Extent

S1 2.71,3.08,3.52 .10,.11,.12 .271, .338,.422

S2 2.43,2.76,3.19 .10,.11,.12 .243,.303,.383

S3 3.03,3.32,3.60 .10,.11,.12 .303,.365,.432


Synthetic Extent Pairs Combinations
S1≥S2* 1
S1 .271, .338,.422 S1,S2
S2≥S1 Formula
S1≥S3 Formula
S2 .243,.303,.383 S1,S3
S3≥S1 1
S2≥S3 Formula
S3 .303,.365,.432 S2,S3
S3≥S2 1

l1-u2/[(m2-u2)-(m1-l1)] V
S2≥S1 S2 .243,.303,.383
l2 m2 u2 .271-.383/
(.303-.383)-(.338-.271) 0.762
S1 .271,.338,.422
l1 m1 u1
S1≥S3 S1 .271,.338,.422
l2 m2 u2 .303-.422/ 0.815
(.338-.422)-(.365-.303)
S3 .303,.365,.432
l1 m1 u1
S2≥S3 S2 .243,.303,.383
l2 m2 u2 .303-.383/ 0.563
(.303-.383)-(.365-.303)
S3 .303,.365,.432
l1 m1 u1

For a combination of S1and S2, check which is greater.


If S1>S2, then value is 1. Then other pair becomes S2>S1for which value is derived by formula
l1-u2/[(m2-u2)-(m1-l1)] where fuzzy matrices are of the form S2(l2,m2,u2), S1(l1,m1,u1)
Degree of possibility values (V)
Pairs Values Criterion Pairs Values Priorities of Priority Normalized
weights d’C1= weights priority
min (val1, val2) weights

S1≥S2 1 S1≥S2
C1 (1,.815) min(1,.815) 0.815 0.343
S2≥S1 .762 S1≥S3

S1≥S3 .815 S2≥S1


C2 (.762,.563) min(.762,.563) 0.563 0.237
S3≥S1 1 S2≥S3

S2≥S3 .563 S3≥S1


C3 (1,1) min(1,1) 1 0.420
S3≥S2 1 S3≥S2

Total 2.378 1
Consistency Index (CI) and Ratio (CR)*

Consistency Index = (λ max – n)/(n-1)


Consistency is closely related to transitive property B>A; A>C = B>C
Consistency Index is compared with random consistency index developed by
Saaty.

n 1 2 3 4 5 6 7 8 9 10
R 0 0 .58 .9 1.12 1.24 1.32 1.41 1.45 1.49
I Ratio = CI/RI ; If CR is less than 10%, then the opinions are
Consistent
consistent.

*Use template to calculate this


Inference
 C3 is considered to be the most important criterion
followed by C1 and C2
 Since consistency ratio is smaller than10%,
evaluation is consistent.
2 level Model – Calculate normalized priority weights for each criterion and sub
criterion separately, multiply each sub criteria weights with respective criteria
weights to derive global priorities
What else?

Compute AHP and FAHP scores, then


 RII (relative Imp Index)
 RMSE
 Agreement Analysis
 Rank ordering to see rank reversal

Das könnte Ihnen auch gefallen