Classification rule discovery using variant genetic algorithm

Abstract

The motive of data mining is to extract information from large database. Classification rule mining is the most used mining technique to acquire hidden knowledge from real world databases for making useful decisions. Obtaining comprehensible rules is very significant in real world applications. In this paper, classification rules have been mined from large volume of database using genetic algorithm based approach. Genetic Algorithm (GA) provides more accurate results than other traditional methods. The primary parameters of GA are crossover, mutation and fitness function. Variations in these operators have shown better impact on accuracy. Experimental results on Fisher's Iris data from UC Irvine (UCI) Machine Learning repository, using proposed variant of GA has shown better performance.