Knowledge Discovery and Data Mining in Medicine - Advanced Techniques in Knowledge Discovery and Data Mining - page 186

true,” “unknown,” “rather false,” “false,” “very false”} to represent the strength of
the connection weight to hidden neurons. The network is trained to make the
connection weights lie in [0,1]. Then the extracted fuzzy rule is represented as
follows:

If (7.9)
where m and n represent the fuzzy linguistic values. In this chapter, we use the
transformation rules shown in Table 7.5, which convert the strength of the
connection weight to the fuzzy linguistic value.
By applying this method to the modified network, we extract fuzzy rules for
the model of the occurrence of hypertension. Table 7.6 shows the fuzzy rules for a
new neuron in the 2nd hidden layer. These rules are not inconsistent with medical
experts' consensus.

x is m true Then y is n true ,

Table 7.5. Truth scales and their correspondin g numerical values.
Linguistic value Numerical value

0 925 . <≤ µ 1 000 .

Very true

0 775 . <≤ µ 0 925 .

True

0 600 . <≤ µ 0775 Rather true

0 .400 <≤ µ 0 600 . Unknown

0 225 . <≤ µ 0 .400

Rather false

0 075 . <≤ µ 0 225 .

False

0 000 . ≤≤ µ 0 075 .

Very false

Table 7.6. The extracted fuzzy rules from a new neuron.
Antecedent part Consequent part

1st SBP is rather true
and 2nd SBP is true
and 3rd SBP is true
and 4th SBP is true
and 5th SBP is true
and 1st DBP is true
and 2nd DBP is true

Predisposition is false

Gender is female
and age(Old) is rather false
and obesity index is rather false
and consumed alcohol is rather false
and 3rd DBP is rather true
and 4th DBP is rather true
and 5th DBP is rather true

Predisposition is true