An improved clustering algorithm based on K-means and harmony search optimization

Clustering is a data mining technique that classifies a set of observations into several clusters based on some similarity measures. The most commonly used partitioning based clustering algorithm is K-means. However, the K-means algorithm has several drawbacks. The algorithm generates a local optimal solution based on the randomly chosen initial centroids. A recently developed meta heuristic optimization algorithm named harmony search helps to find out near global optimal solutions by searching the entire solution space. K-means performs a localized searching. Studies have shown that hybrid algorithm that combines the two ideas will produce a better solution. In this paper, a new approach that combines the improved harmony search optimization technique and an enhanced K-means algorithm is proposed.