Evolving Ensemble Classifier for Mining Data Stream with Concept Drift
Learning concept drift is a challenging task in a non-stationary environment. Concept drift is concerned with learning from data, whose statistical data distribution changes over time. In recent days, ensemble classifiers have become a popular technique and more work has been carried out in data stream classification for non-stationary environment. Ensemble classifiers provide a natural way to adapt the changes which increase the classification accuracy than a single classifier. In this paper, we propose an Evolving Ensemble Classifier (EEC) based on ensemble classification technique which improves the performance of the learning model in the presence of concept drift. The proposed method EEC modifies the weighting function of Accuracy Updated Ensemble (AUE2e) algorithm. Our proposed algorithm EEC is compared with the existing well-known ensemble algorithms such as OzaBaggingo, AWEe and AUE2e on synthetic and real-world datasets. The experimental results show that the accuracy of the proposed algorithm EEC is substantially increased, regardless of the type of concept drift.