Rough fuzzy consistency measure with evolutionary algorithm for attribute reduction

Attribute reduction or feature selection is a mandatory processing part of any pattern recognition, data mining or machine learning system. High dimensional data in a real life problem poses a big challenge for data analysis in real time. Attribute reduction or feature subset selection helps the learning algorithm to perform efficiently by removing irrelevant and redundant information in the data. Feature subset selection requires an efficient measure for evaluation of feature set and an optimal search strategy for finding out the best feature subset from a large number of candidates. Though a large number of algorithms of feature subset selection with various combinations of evaluation measures with search techniques exist, none of them is perfect. Research is still going on in order to find out better algorithm with lesser cost. In this work, a rough fuzzy consistency based measure has been developed for the evaluation of feature set. The measure is combined with genetic algorithm(GA) and particle swarm optimization(PSO), both belong to class of evolutionary computation, for the design of optimal feature subset selection algorithm. Simple simulation experiments with bench mark data set from UCI machine learning database have been done to assess the efficiency of the proposed algorithms and the results are presented in the paper.