HMOSHSSA: a novel framework for solving simultaneous clustering and feature selection problems.

In real-life scenarios, information about the number of clusters is unknown. Due to this, clustering algorithms are unable to generate the valuable partitions. Beside this, the appropriate and optimal number of features is also required to produce the good quality clusters. The selection of optimal number of clusters and feature is a challenging task in the clustering. To resolve these problems, an automatic multi-objective-based clustering approach called HMOSHSSA is proposed in this paper. In HMOSHSSA, the spotted hyena and salp swarm algorithms are hybridized to obtain a better trade-off between these algorithms' intensification and diversification capabilities. Two novel concepts for encoding and threshold setting are incorporated in the HMOSHSSA. The encoding scheme is used to choose the optimal number of clusters and features during the optimization process. The variance of dataset is used for setting the threshold values for both clusters and features. A novel fitness function is proposed to improve the optimization process. The suggested algorithm's performance is evaluated using eight well-known real-world datasets. The statistical significance of HMOSHSSA is measured through t-tests. Results reveal that the proposed approach is able to detect the optimal number of clusters and features from a given dataset without user intervention. This approach is also deployed for solving microarray data analysis and image segmentation problems. HMOSHSSA outperformed the other considered algorithms in terms of performance measures. [ABSTRACT FROM AUTHOR]