Novel Multi-stage Genetic Clustering for Multiobjective Optimization in Data Clustering

Clustering is an unsupervised technique, which partitions the entire input space into regions. These initial partitions have a great impact on the resulting clusters. In this paper, a new Multi Stage Genetic Clustering (MSGC) scheme for multiobjective optimization in data clustering is proposed, which can automatically partition the data into an appropriate number of clusters. K-means is a well-known centroid based algorithm for data clustering and due to a random selection of initial centroids it mostly results into local optima. To overcome this and to get the optimal clusters Genetic Algorithm (GA) is applied at two stages with multiple objective functions. The scheme, proposed in this paper works in two stages for four objective functions. In the sequel, two objective functions in the first stage get the initial partitions, and the other two give optimal cluster centers. The proposed MSGC scheme is implemented for globally accepted datasets, and the results have been compared in terms of Intracluster distance and error rate. The performance analysis shows that MSGC performs significantly better. The results are compared with the existing algorithms to validate our findings. The overall average performance improvement is seen. The average error rate is also minimized, for example, by eight to ten percent for the thyroid dataset as compared to the existing algorithms.