Page 232 - Data Science Algorithms in a Week

216 Abdulrahman Albar, Ahmad Elshennawy, Mohammed Basingab et al.

linguistic terms that are associated with the fuzzy classes. There are a total of 360 rule
assessments in this table, which represents the assessment of 36 rules by ten experts. It is
apparent that 31 of the 36 evaluated rules were evaluated using two or fewer linguistic
terms, and the remaining rules were evaluated with no more than three terms. Five
assessed rules reached full consensus, with an agreement rate of 100%; FLS3-09, FLS3-
20, FLS3-24, FLS3-26, and FLS3-31. It is also observed that twelve assessed rules
received a consensus rate between 80% and 90%, while eighteen rules reached the range
of 60% to 70%. Finally, one rule, FLS3-02, achieved a minimum consensus rate of 50%.
The average consensus rate for this subsystem is 76%, which when compared to the
average rate of 79% for subsystem I, is relatively close, even though subsystem III
featured more inputs. When compared to subsystem II’s average consensus rate of 84%,
76% is still satisfactory, although subsystem III contained more assessment classes. The
frequency of linguistic class use in assessing rules was the highest in the “high” class
with 124 uses, followed by “medium” with 105 uses, while the least used classes were
“low” and “very high”, with 66 and 65 uses, respectively.

Table 9: Results of expert evaluation for subsystem III’s fuzzy rules.

The final list of fuzzy rules for subsystem III is provided in Appendix C, which will
be stored in the fuzzy rule base to build the fuzzy knowledge base.
The results for subsystem IV’s rule assessments are provided in Table 10, which is
the most significant subsystem in the fuzzy system. In this subsystem, ten experts
evaluated 80 rules against five assessment levels, with each rule consisting of a
combination of three AND conditions. As each rule is designed with three combinations
for the antecedent, to be assessed at five levels, this subsystem presents the highest
complexity for expert assessment.