An Ensemble Multi-Model Voting Method for Adapting to Concept Drift

Aiming at the challenges posed by concept drift in streaming data mining, this paper proposes an Ensemble Multi-Model Voting Method for Adapting to Concept Drift (EMVM_ATCD). The method employs integrated multi-classifiers to improve model stability, uses online learning methods to update the model, and adds a dropout layer to force the model to learn different combinations to enhance generalization ability. A voting mechanism is used to process the model prediction results to enhance the ability to cope with concept drift. Experimental results show that the method achieves performance improvements ranging from 0.1% to 10% on multiple datasets, proving that it can effectively handle various types of data.