Rule Extraction from Trained Neural Network with Evolutionary Algorithms

Abstract

This paper describes a solution to the problem of incomprehensibility of the neural network by introducing simultaneously working Evolutionary Algorithms as a tool for extracting set of rules in the form of if — then. Each Evolutionary Algorithm is working for searching rules describing one class, which is recognized by a Neural Network. The proposed method has been tested on real domains in order to analyze its behavior under various conditions. A comparison with other rule extraction methods is presented as well.