An ensemble classifier method for classifying data streams with recurrent concept drift

In order to solve the problem that existing ensemble classifier algorithms can´t recognize the recurrent concept drift effectively, a new algorithm called Historical Classifier Ensembles for classification (HCE) is proposed. By storing the historical classifiers and ensembles, the algorithm can make full use of the historical concept information and can improve the classification efficiency and accuracy in data stream classification with concept drifts. The experiment results show that the HCE algorithm adapts better to data streams environment with implied recurrent concept drifts.