Incremental Clustering for the Classification of Concept-Drifting Data Streams

Abstract. Concept drift is a common phenomenon in streaming data environments and constitutes an interesting challenge for researchers in the machine learning and data mining community. This paper proposes a probabilistic representation model for data stream classification and investigates the use of incremental clustering algorithms in order to identify and adapt to concept drift. An experimental study is performed using three real-world datasets from the text domain, a basic implementation of the proposed framework and three baseline methods for dealing with drifting concepts. Results are promising and encourage further investigation. 1