Data clustering using enhanced biogeography-based optimization

Abstract

Data clustering is one of the important tool in data analysis which partitions the dataset into different groups based on similarity and dissimilarity measures. Clustering is still a NP-hard problem for large dataset due to the presence of irrelevant, overlapping, missing and unknown features which leads to converge it into local optima. Therefore, this paper introduces a novel hybrid meta-heuristic data clustering approach which is based on K-means and biogeography-based optimization (BBO). The proposed method uses K-means to initialize the population of BBO. The simulation has been done on eleven dataset. Experimental and statistical results validate that proposed method outperforms the existing methods.