A Black Widow Optimization Algorithm for Clustering Problems

Clustering methods that have been implemented in many fields of science and technology have data points for every cluster that shares some or all properties. k-means algorithm is used for data clustering. In our problem, the clustering method is heuristic and may be stuck in local optima; so, we propose a black widow optimization algorithm (BWOA) for data clustering. The fitness of the clustering can be calculated by using the sum of Euclidian distances of every data point from its cluster center vector. The experimental data sets called Iris and Seeds are used to prove our claim. The results show that the proposed BWOA generates a higher accuracy equals 96.67% for the Iris data set and generates a higher accuracy equals 92.86% for the Seeds data set with higher clustering accuracies better than those formerly reported in the literature.