Parallel K-Means Clustering For Large Data Sets

Large data set clustering is a time-consuming task that can be improved by parallel techniques. One of the most popular clustering algorithms is K-Means. This algorithm simply selects K points as cluster centers and assigns each data point to its nearest center. The Algorithm reassigns the cluster centers until a convergence criterion would be met. In this paper, a new method for parallelizing the K-Means is presented. In the proposed algorithm, each cluster is placed in a separate site. This property can be useful for applications that need to store and process similar data (each cluster) separately. In addition, against very other methods it does not make any assumptions about the homogeneity of data distribution. The experimental results show that the proposed algorithm speeds up the clustering process significantly

Keywords:

Parallel clustering , K-Means , cluster center