An intelligent hybrid feature selection using correlation coefficient and particle swarm optimization on microarray gene expression data

Abstract

In this paper author is describing concept to apply combination of Particle Swarm Optimization algorithm and Correlation Coefficient algorithm for hybrid features selection to increase classifier accuracy and decrease system execution time. Some datasets such as GENES may contain attributes in thousands and classifying such huge attributes may degrade classifier accuracy and increase system execution time. To overcome from this issue author is using PSO and Correlation algorithm to select important attributes from dataset and ignoring unimportant attributes. These feature selection algorithms will prune unrelated attributes and select few attributes to perform classification. In this paper we are using ‘Lymphoma’ genes dataset which contains more than 4000 attributes but by applying PSO it will select only 52 important attributes out of 4000. Another dataset called ‘SRBCT’ contains more than 2000 attributes but PSO will choose few attributes from 2000. The main aim of this project is to select features from dataset by applying PSO features selection algorithm to reduce dataset size.