A Framework of Cluster Decision Tree in Data Stream Classification

Recently, data streams classification with concept drifting has drawn increasing attention of scholars in data mining, due to the deficiencies of existing algorithms in accuracy and efficient. In this paper, we propose a framework for handling the problem mentioned above using cluster decision tree. We cluster those data which cannot be classified temporarily into n class, and generate new branches of the VFDT based on cluster result or replace original ones. Our empirical study shows that the proposed method has substantial advantages over traditional classifiers in prediction accuracy and efficiency.