Application of Gravitational Search Algorithm on Data Clustering

Data clustering, the process of grouping similar objects in a set of observations is one of the attractive and main tasks in data mining that is used in many areas and applications such as text clustering and information retrieval, data compaction, fraud detection, biology, computer vision, data summarization, marketing and customer analysis, etc. The well-known k-means algorithm, which widely applied to the clustering problem, has the drawbacks of depending on the initial state of centroids and may converge to the local optima rather than global optima. A data clustering algorithm based on the gravitational search algorithm (GSA) is proposed in this research. In this algorithm, some candidate solutions for clustering problem are created randomly and then interact with one another via Newton’s gravity law to search the problem space. The performance of the presented algorithm is compared with three other well-known clustering algorithms, including k-means, genetic algorithm (GA), and particle swarm optimization algorithm (PSO) on four real and standard datasets. Experimental results confirm that the GSA is a robust and viable method for data clustering.