Mining Concept-Drifting and Noisy Data Streams Using Ensemble Classifiers

Mining concept drifting data stream is a challenging area for data mining research. Recent years have witnessed an averaging ensemble classifier which is based on the learnable assumption, although this ensemble classifier is an efficient algorithm for mining concept-drifting data streams, it is still inadequate to represent real-world data streams with noisy data. In this paper, we propose a novel ensemble classifier framework for mining concept-drifting data streams with noise. The method, called WEAP-I, which trains a weighted ensemble classifier on the most n data chunks and trains an averaging ensemble classifier on the most recent data chunk. All the base classifiers are combined to form the WEAP-I ensemble classifier. Our theoretical and empirical study shows that our framework is superior and more robust to averaging ensemble for noisy data streams.