Parallelization of K-Means Clustering on Multi-Core Processors

Multi-core processors have recently been available on most personal computers. To get the maximum benefit of computational power from the multi-core architecture, we need a new design on existing algorithms and software. In this paper we propose the parallelization of the well-known k-means clustering algorithm. We employ a single program multiple data (SPMD) approach based on a message passing model. Sending and receiving messages between a master and the concurrently created process are done in an asynchronous manner. Therefore, the implementation can be highly parallel and fault tolerant. The experimental results demonstrate considerable speedup rate of the proposed parallel k-means clustering method, compared to the serial k-means approach. Key-Words: Parallel k-means, Multi-core processing, Concurrent programming, Erlang, Functional language, Clustering