Distributed single pass clustering algorithm based on MapReduce

Abstract

Available data increase quickly every moment, this eventually drags to big data flooding. Hence there is an emergent need for exploiting big data in order to extract valuable knowledge from it. Adoption of distributed architecture and data intensive algorithms facilitates handling and processing big data. This paper introduces a distributed single pass clustering algorithm based on MapReduce in order to reduce running time of processing big data. Also, it introduces median based single pass clustering in order to mitigate the order of the input data problem that is associated with single pass clustering. Furthermore, it introduces a new hybrid approach which integrates median based single pass clustering and k-means algorithm. The proposed integration improves the median based clustering to work well with sparse data such as text. The experimental results state that the proposed approaches outperform traditional single pass clustering.