Hierarchical Density-Based Clustering Using MapReduce

Hierarchical density-based clustering is a powerful tool for exploratory data analysis, which can play an important role in the understanding and organization of datasets. However, its applicability to large datasets limited because computational complexity hierarchical methods has quadratic lower bound number objects be clustered. MapReduce popular programming model speed up mining machine learning algorithms operating on large, possibly distributed In literature, there have been attempts parallelize such as Single-Linkage, principle also extended broader scope clustering, but are inherently difficult with MapReduce. this paper, we discuss why adapting previous approaches Single-Linkage using leads very inefficient solutions when one wants compute hierarchies. Preliminarily, solution, based exact, yet computationally demanding, random blocks parallelization scheme. To able efficiently apply MapReduce, then propose different scheme that computes approximate hierarchy much faster, recursive sampling approach. This approach HDBSCAN*, state-of-the-art algorithm, combined summarization technique called bubbles. The proposed method evaluated terms both runtime quality approximation datasets, showing effectiveness scalability.