Using ART networks with rule extraction as a data mining tool

Abstract

In this paper, two important mining tools, i.e. Neural Networks and Genetic Algorithm (GA), have been used for mining the database through pattern classification. The processing methodology consists of three major phases: Network construction and training; Pruning; Rule extraction and validation. The neural networks used are: Adaptive Resonance

Theory 1.5<i> </i>(ART1.5),<i> </i>Adaptive Resonance Theory3 (ATR3) and Multi Channel ART (MART).The pruning phase aims at removing redundant links and units without increasing the classification error rate of the network. The pruning methods used are: Local Pruning and Threshold Pruning. The final phase extracts the classification rules from the final weights of the pruned network in the form of IF-THEN rules and rules are validated using validation algorithm. The genetic algorithm is also used to optimise the performance with regard to the various parameters used in the network. The proposed networks have been tested on letter recognition data taken from UCI machine learning site. Simulation algorithms have been implemented using C++ language in the Pentium machine. The graphical results have been obtained using Micro Soft Excel.