Parallel Black Hole Clustering Based on MapReduce

One of the key reasons that traditional clustering methods are inefficient for analyzing large-scale datasets is because most of them are designed for a centralized system. This means that if the size of input data exceeds the size of storage or memory of such a system, it would make the task of clustering much more difficult. To mitigate this problem, an efficient clustering algorithm, called MapReduce Black Hole (MRBH), is presented in this paper to leverages the strength of the black hole algorithm and the MapReduce programming model of Hadoop to accelerate the clustering speed by both software and hardware. By using MapReduce, MRBH will then divide a large dataset into a number of small data sets and cluster these smaller data sets in parallel. Moreover, MRBH inherits the characteristics of the black hole algorithm, meaning that no parameters are to be set manually, thus, the implementation is easy. To evaluate the performance of the proposed algorithm, several datasets are used with different numbers of nodes. Experimental results show that the proposed algorithm can provide a significant speedup as the number of nodes increases.