Extracting classification rules using neural networks and gene expression programming

Abstract EN

In this paper, we propose a method to extract classification rules from data base by using neural networks and Gene Expression Programming (GEP).

A constructive learning algorithm of three layered feed forward neural network is proposed to train the network.

The procedure extracts rules from the pruned network by clustering the hidden units activation values in order that the hidden-unit activation values are restricted to discrete values.

Hence, the output of the hidden layer and network’s output can be easily interpreted in the form of rules (ho-¬rules).

Due to the large number of examples entered to the network, it is difficult to extract the rules between the hidden layer and input layer.

Hence, GEP is employed to solve this problem by encoding the rule between output and input layer in the form of logic expression.

The computational results have shown that the system extracted a very compact, comprehensible rule set without unduly reducing the accuracy rate, in comparison with the accuracy rate of the rule set discovered by the well-known c4.5 algorithm and neural networks via Genetic Algorithms.