Improved Particle Swarm Optimization Based K-Means Clustering

Clustering is a popular data analysis and data mining technique. K-Means is one of the most popular data mining algorithms for being simple, scalable and easily modifiable to a variety of contexts and application domains. The major issue of traditional K-Means algorithm is that its performance depends on the initialization of centroid and requires the number of clusters to be specified in advance. Many evolutionary based clustering algorithms have been developed in recent years for selecting optimum initial centroid to optimize clustering results. Particle Swarm Optimization algorithm is a population-based memetic-evolution-motivated meta-heuristic algorithm that mimics the capability of swarm. The K-Means algorithm typically uses Euclidean or squared Euclidean distance to measure the distortion between a data object and its cluster centroid. The Euclidean and squared Euclidean distances are usually computed from raw data and not from standardized data. Normalization is one of the important preprocessing steps, to transform values of all attributes. Effective data clustering can only occur if an equally effective technique for normalizing the data is applied. This paper proposes an effective partitional clustering algorithm which is developed by integrating the merits of Particle Swarm Optimization and normalization with traditional K-Means clustering algorithms. Experiments are conducted on real dataset to prove the efficiency of the proposed algorithm.