multi-label data stream classification using Heterogeneous ensemble learning

In many real-world applications huge amounts of data are continuously generated in the form of data streams. Multi-label data stream classification is a methodology that tries to solve data stream classification problems where multiple classes are associated with each data sample. Changes in data distribution, also known as concept drift, cause existing data stream classification models to rapidly lose their effectiveness. Ensemble classifiers has been widely applied to data stream classification and heterogeneous ensemble classifiers by combining several types of different learning models can achieve greater diversity among its members, which helps to improve its performance. In this study, we introduce a heterogeneous ensemble method for multi-label data streams Classification and propose a concept drift detection algorithm that exploits label correlations and dependencies between class labels for multi-label data streams. Experiments show that the proposed method achieves competitive performance in comparison to other benchmark methods.

کلیدواژه ها:

Multi label data stream ، classification ، concept drift ، heterogeneous ensemble ، label correlations