Unsupervised Concept Drift Detection Using Dynamic Crucial Feature Distribution Test in Data Streams

Distribution of data often changes over time and leads to the unpredictable changes in the implicit information behind data streams. This phenomenon is referred to as Concept Drift. The accuracy of conventional models reduces as time goes by, and old models are rendered impractical. In this paper, we propose a novel approach for solving the concept drift detection problem using the unsupervised method and focusing on the dynamic crucial feature distribution test. Extensive experiments have been done to evaluate the performance of the proposed method against classic and state-of-the-art methods. Experimental results demonstrate the efficacy of the proposed model when applied to synthetic as well as real-world datasets.