Automatic Clustering Using a Genetic Algorithm with New Solution Encoding and Operators

Genetic algorithms (GA) are randomized search and optimization techniques which have proven to be robust and effective in large scale problems. In this work, we propose a new GA approach for solving the automatic clustering problem, ACGA Automatic Clustering Genetic Algorithm. It is capable of finding the optimal number of clusters in a dataset, and correctly assign each data point to a cluster without any prior knowledge about the data. An encoding scheme which had not yet been tested with GA is adopted and new genetic operators are developed. The algorithm can use any cluster validity function as fitness function. Experimental validation shows that this new approach outperforms the classical clustering methods K-means and FCM. The method provides good results, and requires a small number of iterations to converge.