Page 221 - Data Science Algorithms in a Week

Managing Overcrowding in Healthcare using Fuzzy Logic 205

 List all possible permutations of “AND” rules for each fuzzy logic subsystem.
 Code each rule with “FLSm-n” where FLS stands for Fuzzy Logic Subsystem, m
stands for the number of subsystem, and n stands for the rule number within the m
subsystems.
 Code “HCE-k” for each participating expert, where HCE stands for Healthcare
Expert, and k stands for the expert number.
 The Expert HCE-k determines the consequence of the fuzzy conditional statement
FLSm-n based on their expertise.
 The fuzzy conditional statement FLSm-n must meet a 50% consensus rate among
experts, and must be the only consequence to receive a 50% consensus rate, to be
accepted as a valid fuzzy rule.
 If the consensus rate does not meet the determined criteria, further iterations should
be conducted with a new expert until the consensus rate achieves the criteria in the
previous step.

The process for developing fuzzy rules is illustrated in Figure 10, where the
consensus feedback is elaborated upon in more detail.

Figure 10: Process for developing Fuzzy Rules.

Fuzzification Process

Fuzzification is the first step in the fuzzy system, as it obtains both the membership
function type and the degree of membership from the database. This database is built
from the surveyed expert determination of membership function intervals. In the
fuzzification process, crisp values which are within the universe of discourse of the input
variable are translated into fuzzy values, and the fuzzifier determines the degree to which
they belong to a membership function. The fuzzifier for this designed fuzzy system
adapts the Minimum approach. Whereas the input is crisp, the output is a degree of
membership in a qualitative set. The fuzzified outputs allow the system to determine the
degree to which each fuzzy condition satisfies each rule.