An improved data clustering algorithm in a multiobjective framework

Cluster analysis is an important step in data mining. For clustering, various multiobjective techniques are evolved, which can automatically partition the data into an appropriate no. of clusters. K-means is a well known data clustering algorithm and is proven to be better for many practical applications. The proposed work is based on achieving multiple objective functions for data clustering thereby, improving the quality. To achieve this, the K-means algorithm is used for producing the initial clusters. These clusters are then optimized by using three objective functions as a fitness function in the NSGA II algorithm. Three objective functions such as compactness, connectedness, and symmetry of the cluster are optimized simultaneously. The results are compared with the existing multiobjective algorithms and a significant improvement is observed.