An Augmented Learning Approach for Multiple Data Streams Under Concept Drift

Multiple data streams learning attracts more and more attention recently. Different from learning a single data stream, the uncertain and complex occurrence of concept drift in multiple data streams, bring challenges in real-time learning task. To address this issue, this paper proposed a method called time-warping-based concept drift learning method (TW-CDM) for dealing with multiple data streams. First, a time-warping-based drift identification process is given to recognize the drift region. Second, an augmented learning process is developed by crossly using the located region data. Finally, a selectively augmented learning process is given to reduce the influence of different drift severity. The proposed method is evaluated on both synthetic and real-world datasets, and compared with benchmark methods. The experiment results show the efficiency of the proposed method.