Enhancement of Data Streaming in Clustering for Uncertain Data

Abstract

Data stream mining presents exciting model from the stream of information. The objective of clustering data stream is to cluster and determine unknown pattern from streaming data. Streaming data generated from imprecise hardware result into uncertainty built into them. In this paper, we propose a novel approach for data stream clustering which reduces the degree of uncertainty and increases the degree of homogeneity. Unlike current procedures whose aim is to cluster data stream, we consider the case where uncertainty of micro-cluster affects the quality of clustering outcome. The proposed algorithm Fuzzy-HCMStream is based on HCMStream algorithm which incorporates the concept of Fuzzy C-Mean algorithm, in account for uncertainty. Our proposed algorithm increases degree of homogeneity, which leads to improvement in clustering quality. The implementation of proposed algorithm is tested with two data sets comprising synthetic and real data set. The Fuzzy-HCMStream algorithm shows improvement in performance compared to the other state-of-the-art algorithms in terms of various indices for evaluating clustering results.