Indian Journal of Science and Technology

Abstract

Objective: To design automatic data clustering algorithm that find number of clusters automatically with balance between exploration and exploitation search space. Methods: This work proposes Spider monkey optimisation with tabu search algorithm named as SMOTS for automatic data clustering. In this algorithm, the local search of spider monkey is improved with tabu search algorithm. For better results, compact separated index with Gaussian kernel distribution is introduced as a fitness function. The experiments are performed on Vowel, Iris, Wine, Seed, E.coli and Thyroid data sets. The results are validated with cluster optimality, inter and intra cluster distances with 5 well known and 7 recently published algorithms like DE, GA, GWO, WOA, PSO, AHPSOM, ACICA, ACDCSA, AC-MeanABC, TMKGSO, Black Hole k-Means, and EOAK-means. To test the statistical significance of the proposed algorithm an unpaired t-test is performed between SMOTS and second best algorithms on mean inter cluster distance. Findings: In comparison with well-known clustering algorithm on six data set SMOTS produced 100%, 33.33%, 83.33% accurate results on cluster optimality, intra and inter cluster distance respectively. In comparison with recently published algorithms on six data set SMOTS produced 50%, 66.66%, 50% accurate results on cluster optimality, intra and inter cluster distance respectively. The hypothesis testing results shows that p-value of the t-test is less than 1% except vowel data set means SMOTS is highly statistically significant compare to second best algorithms. Novelty: In real life data set information about number of cluster is rarely available and this produced faulty results. Proposed method can process data without any prior information of number of clusters and data distribution with accurate results. Keywords: Spider Monkey Optimisation; Tabu Search; Automatic Clustering; Neighbour Search; Swarm Intelligence 2