Data Clustering Using Group Search Optimization with Alternative Fitness Functions

Data clustering is an important tool for statistical data analysis and exploration, and it has been successfully applied in many fields like image understanding, bioinformatics, big data mining, and so on. From the past few decades, Evolutionary Algorithms (EAs) have been introduced to deal with clustering task, given their global search capabilities and their mechanisms to escape from local minima points. EAs execution is driven in an attempt to optimize a criterion function, also known as fitness function. In this work, we evaluate the influence of the fitness function on Group Search Optimization (GSO) meta-heuristic when applied to data clustering. Three different fitness function are proposed to GSO. Experiments are performed on twelve benchmark data sets obtained from UCI Machine Learning Repository to evaluate the performance of all alternative GSO models in comparison to other well-known partitional clustering methods from literature.