Cluster Pattern Matching Using ACO Based Feature Selection for Efficient Data Classification

Abstract

Cluster Pattern Matching based Classification (CPMC) is a classification technique based on a similarity measure between the training instances and the unknown sample. An Ant Colony Optimization based feature selection is proposed to select the features. According to this approach, the training data set is clustered. The cluster to which the unknown sample belongs is found and each of the selected features of the unknown sample is compared with the corresponding feature of the training instances in the cluster and the class of the unknown sample is predicted based on majority voting of class labels having highest number of matching patterns. A probabilistic approach is used to predict the class label when more than one class label has the same majority. Experimental results demonstrating the efficiency of classification accuracy of CPMC are shown to prove that the proposed approach is better when compared to existing classification techniques.