Simultaneous Clustering and Feature Selection using Social Group Optimization with Dynamic Threshold Setting for Microarray Data

In this research, a unique method for automatically and simultaneously choosing significant features as well as cluster numbers from a dataset is proposed. Social Group Optimization (SGO) algorithm is used as metaheuristic. The SGO incorporates two new ideas for threshold setting and encoding. During the optimization phase, a number of features and cluster centers are encoded using the encoding scheme. The dataset variance is utilized to determine the value of threshold for both clusters as well as features. A new clustering criterion is employed to enhance the efficiency of the search process. We compare the proposed algorithm's performance to eight freshly developed clustering algorithms and evaluate it on nine well-known real-world datasets. The statistical significance of the SGO clustering technique is determined using T-tests. The outcomes demonstrate that the proposed method can optimally identify the number of clusters as well as features from a dataset without any input from the user. In order to demonstrate the algorithm's accuracy and success, microarray data is also analyzed using this method.