C-LAS Relief-An Improved Feature Selection Technique in Data Mining

Abstract

selection or Feature subset selection is a process of reducing the attribute space in the feature set. It is also stated that feature selection is a technique of identifying a subset of features. These subsets of features are selected by removing irrelevant or redundant features in the feature set. A good feature set is said to be that it contains highly correlated features with the class. Such feature set improves the efficiency of the classification algorithms and also the classification accuracy. The Chebyshev distance with median variance in the weight estimation of attributes in the Relief imparts the consistency and good accuracy. In this paper a novel algorithm called C LAS-Relief is used to improve the reliability and accuracy of classification. Here C LAS-Relief stands for Chebyshev distance LAS-Relief. The efficiency and effectiveness of proposed method is experimented using agriculture soil data sets, Soybean and Ozone data sets. Similarly the new approach is compared with LAS-Relief approach using Naive bayes and J48 classifiers. The classification accuracy of C-LAS-Relief is superior over LAS- Relief. C LAS-Relief algorithm increases the accuracy of classification compared to LAS-Relief algorithm.