MapReduce-based Dragonfly Algorithm for large-scale Data-Clustering

The improvement in networking-technology, sensors, and availability of the internet have accelerated the huge growth to the electronic data. This immense amount of data has stimulated the development of new data analysis methods for better decision making. Clustering is an influential unsupervised data analysis approach, with wide areas of applications. K-Means is a fast and prolific approach for data-clustering present in the literature. However, the algorithm is easily-influenced by the initial positions of cluster-centroids, and the method converges to the local optimum that is nearest to the initial centroid positions. Furthermore, the algorithm cannot process large datasets within a reasonable time-period. In this work, a novel data-clustering method named MapReduce-based Dragonfly Algorithm (MR-DA) is introduced. The efficiency of MR-DA is compared with 4 other recent methods. The experimental results demonstrate that MR-DA surpassed the other considered methods on the majority of the datasets.