Model-Building Techniques of the Generalized Characteristic Expert Evaluation Models - Expert Fuzzy Information Processing - page 141

Table 4.5 Elements of a matrix and transitive closure of the expert models pairwise
consistency matrix

Matrix elements

1 0,819 0,875 0,923 0,792

0,819 1 0,812 0,834 0,768

0,875 0,812 1 0,938 0,884

0,923 0,834 0,938 1 0,858

0,792 0,768 0,884 0,858 1

Elements of transitive closures

1 0,834 0,923 0,923 0,884

0,834 1 0,834 0,834 0,834

0,923 0,834 1 0,938 0,884

0,923 0,834 0,938 1 0,884

0,884 0,834 0,884 0,884 1

Indexes of pairwise consistency of expert models (Table 4.5) were calculated
according to the definition (see §3.1).
As one can see from this matrix, models of the third and fourth experts have the
greatest index of pairwise consistency.
Further, the fuzzy binary relation of similarity based on the computed indexes
of pairwise consistency of expert models is constructed. The matrix of pairwise
consistency of expert models is not transitive; therefore its transitive closure
(see the bottom part of Table 4.5) is determined.

Table 4.6 Fuzzy clusterization of expert models under similarity relation ˆ

Confidence level Cluster

{1,2,3,4,5}

0,834

0,884 {1,3,4,5},{2}

0,923 {1,3,4},{2},{5}

0,938 {3,4},{1},{2},{5}

1 {1},{2},{3},{4},{5}

This matrix defines using set of expert models the fuzzy relation of similarity
ˆ with fuzziness degree equal to 0.196. Relation ˆ is decomposed onto
equivalence relations, the results obtained are given in Table 4.6, and weight
coefficients (see §4.2) of expert models (Table 4.7) are determined based on the
similarity R ˆ .
Fuzziness degrees of expert models for software completeness evaluations are
determined:

ζ = 0 30 ζ = 0 30

ζ = 0 30

ζ = 0 26 ζ = 0 32

, , , , .

3 5

1 4

2