Sustainable automatic data clustering using hybrid PSO algorithm with mutation

Widespread use of various mobiles, social networks and IOT devices results into continuous generation of the data, often leading to the formation of the big data. Sustainable grouping of such data into various clusters is an open research problem, which aims to provide solutions which are computationally efficient and maintainable over dynamic data. This paper proposes a new sustainable clustering algorithm HPSOM by hybridization of PSO with mutation operator for clustering of the data generated from different networks. The data generated by such networks is usually dynamic and heterogeneous in nature and the number of clusters is not fixed/ known in advance. Hence the proposed algorithm is further extended as AHPSOM for generating and re-adjusting the clusters automatically over the mobile network devices, and it facilitates the generation of sustainable clusters. Firstly, the performance of basic HPSOM is evaluated on six real life data sets and is also compared with some known evolutionary clustering techniques in terms of SWCD and Convergence speed. Then the performance of AHPSOM is evaluated using some synthetic and real life datasets using some validity metrics (cluster numbers, intra cluster distance, inter cluster distance, ARI and F-measure) and is also compared with some prevalent state-of-art automatic clustering techniques. The results show that the proposed algorithm is very efficient in terms of creating well separated, compact and sustainable clusters. (C) 2019 Elsevier Inc. All rights reserved.