Simultaneous Clustering and Feature Selection Using Nature-Inspired Algorithm

Abstract

Clustering means partitioning of data set into individual clusters where the similarity among the data exists and the procedure requires more substantial methodology when the dimension of the input data set is very high as well as we have to select more relevant dimensions or features which are necessary enough for clustering. Nature-inspired algorithm like firefly gives a promising result in function optimization and clustering. The proposed work will represent a new feature selection cum clustering algorithm called iterative firefly k-means features selection (FKM_FS) algorithm by minimizing the inter-cluster distance as well as maximizing the intra-cluster distance and maximizing the average relevance of the particular feature to the clustering. We define a methodology based on variance of observation in a cluster with respect to global variance to identify relevant feature subset. Finally, the algorithm will run both on real and data set.