Improving the efficiency of data clustering with chaotic evolutionary algorithms

Nowadays, clustering plays an important role in most research fields such as engineering, medicine, biology, data mining, etc. In fact, clustering means unsupervised division. By using it, the data are divided into categories that are more similar to each other in terms of the parameters of interest. One of the famous methods in this field is k-means. In this method, despite the dependence on initial conditions and convergence to local optimal points, N numbers of data are grouped into k clusters with high speed. In this article, to solve the existing problems, the combined method is used based on evolutionary algorithms, chaos theory and k-means; that is in addition to solving the mentioned problems, it will also be independent of the number of variables. In this article, for the purpose of validation, the proposed methods are implemented on 13 different famous collections, and the results are compared with genetic algorithm, particle community, bee colony, simulated refrigeration, differential evolution, harmony search, and k-means methods. The high ability and robustness of these methods will be evident based on the results.