Research Article
Tuning Expert Systems for Cost-Sensitive Decisions
Algorithm 1
Expert system tuning algorithm.
Expert_System_Tuning (, , , ) | Input: | : An expert system. Given the input vector of a problem case, suggests a CF value. By | default, the expert system makes a positive classification decision if the CF is positive and a | negative classification decision, otherwise. | : A sample of solved cases , each of which is a pair , where | and . In addition, let denote the CF value suggested | by the expert system for the case . | : Cost ratio. . | : Prior probability of the positive class. | Output: A tuned cutoff for the CF value. | BEGIN | := := 0. | FOR := 1 TO , | Classify using the expert system, that is, := . | IF , THEN | := + 1. | ELSE | := + 1. | Sort the cases in on the CF value in ascending order. | := . | FOR := 0 TO , | IF , THEN | := := −1. | ELSE IF , THEN | := . | := . | ELSE | := := 1. | IF , THEN | := ; := 0; | ELSE IF , THEN | := + 1. | ELSE | := – 1. | Assuming the cutoff of , compute the expected misclassification cost as | . | IF , THEN | := ; := ; := . | RETURN . | END. |
|