Mining of data stream using “DDenStream” clustering algorithm

Abstract

Many real time applications, they are generated continues flow of data streams have became more popular now a days. Therefore many researches attracted clustering data streams. Most of data stream clustering algorithms based on distance function which find out clusters with spiracle of shape clusters and unable to deal noisy data. Therefore density based clustering algorithms substitute remarkable solution to find out clusters with arbitrary shapes of cluster and also handling noisy data. In which the clustering is performing with the bases of high density area of objects and also segregate low density objects as noise. In this paper we studied a simple existing data stream clustering algorithm DenStream based on DBScan. Based on DenStream a novel data stream clustering approach “DDenStream” is proposed. DDenStream is a modified data stream clustering algorithm of DenStream. It is based on fading window model therefore it applies a density decaying technique when the evolving data streams are captured and improved the quality of clusters by extracts the boundary points when two or more than microclusters are overlapping each other by using DCQ-Means algorithms. This approach also resolved the overlapping problem of microclusters. It find out arbitrary specs and good quality of clusters with noise.