Article Preview
Top1. Introduction
Cluster analysis is an unsupervised machine learning approach, which is used to partition dataset into different subsets (Kushwaha et al., 2018; Kant & Ansari, 2016). The data items associated with same subset are more similar in nature than other (Aggarwal & Reddy, 2013). Broadly, clustering methods are divided into two categories: hierarchical and partitional clustering (Xu & Wunsch, 2005). Hierarchal clustering is also divided into two groups i.e. agglomerative and divisive. The agglomerative method works in bottom up fashion, whereas, divisive method works in top down manner. In partitional clustering method, data is divided into several disjoint clusters that are optimal in nature. It is observed that large number of meta-heuristic algorithms has been reported in literature to find optimal results for clustering problems (Shelokar et al., 2004; Xiao et al., 2010; Cura, 2012; Taherdangkoo et al., 2013; Kumar & Sahoo, 2014; Bahrololoum et al., 2015). It is also seen that several improved versions of these meta-heuristic algorithms are also presented to solve clustering problems effectively (Chuang et al., 2011; Jiang & Wang, 2014). Moreover, some hybrid versions of meta-heuristic algorithms also developed to determine effective clustering results as well as to overcome shortcomings of existing clustering algorithms (Abualigah et al., 2017; Sheng et al., 2010; Huang et al., 2013; Yan et al., 2012; Bouyer & Hatamlou, 2018; Kumar & Singh, 2018; Wang et al., 2016). It is noticed that several issues are associated with the performance of clustering algorithms like local optima, population initialization and convergence speed etc. (Cao et al., 2009; Kang et al., 2016; Kumar & Singh, 2019). To overcome these issues, researchers either hybridized the existing algorithms or developed new algorithm to obtain better clustering results.