Data Clustering using Differential Search Algorithm

The main challenges of clustering techniques are to tune the initial cluster centres and to avoid the solution being trapped in the local optima. In this paper, a new metaheuristic algorithm, Differential Search (DS), is used to solve these problems. The DS explores the search space of the given dataset to find the near-optimal cluster centres. The cluster centre-based encoding scheme is used to evolve the cluster centres. The proposed DS-based clustering technique is tested over four real-life datasets. The performance of DS-based clustering is compared with four recently developed metaheuristic techniques. The computational results are encouraging and demonstrate that the DS-based clustering provides better values in terms of precision, recall and G-Measure.