Different aspects of data stream clustering

Abstract

Nowadays the growth of the datasets size causes some difficulties to extract useful information and knowledge especially in specific domains. However, new methods in data mining need to be developed in both sides of supervised and unsupervised approaches. Nevertheless, data stream clustering can be taken into account as an effective strategy to apply for huge data as an unsupervised fashion. In this research we not only propose a framework for data stream clustering but also evaluate different aspects of existing obstacles in this arena. The main problem in data stream clustering is visiting data once therefore new methods should be applied. On the other hand, concept drift must be recognized in real-time. In this paper, we try to clarify: first, the different aspects of problem with regard to data stream clustering generally and how several prominent solutions tackle different problems; second, the varying assumptions, heuristics, and intuitions forming the basis of approaches and finally a new framework for data stream clustering is proposed with regard to the specific difficulties encountered in this field of research. © 2013 Springer Science+Business Media.