Ensemble concept drift detection in data stream mining: A Review

In the data-driven era, machine learning plays a vital role in analyzing and processing big data. One of the fundamental challenges in this area is managing conceptual drift in data streams, where the changing distribution of data reduces the accuracy of learning models and makes them ineffective in predicting the future . Traditional classifiers are not expected to learn patterns in non-stationary distributions of data. For any real-time use, the classifier must detect concept drift and adapt over time. Compared with concept drift detection for a data stream, the challenges of ensemble concept drift detection arise from three aspects: first, the training data becomes more complex, Second, the underlying distribution becomes more complex , and third, the correlation between data streams becomes more complex . In this article, we provide a comprehensive review of ensemble concept drift detectors in data stream mining, and also review their techniques, key points, advantages, and limitations .