Empirical studies on applying density-based clustering to stream data

Abstract

Density-based clustering has advantages over partition-based clustering, such as K-means, in that it does not need to specify the number of clusters (k) and can generate clusters of arbitrary shape. However, density-based clustering requires hyper-parameters such as proximity distance and the minimum number of proximity data that are suitable for data characteristics, and this greatly influences the clustering performance. In this paper, we present a density-based clustering algorithm for stream data, which exploits coresets in the sliding window model. We provide an experimental analysis of these hyper-parameters on the performance of the algorithm.