CWIN: A New Windowing Technique for Detecting Concept Drift in Data Streams
Ranjan Bhat, Sharanya Prabhu, Neelima Bayyapu · 2023
Due to the dynamic nature of data streams, concept drift detection is a crucial feature for any live data analytics algorithm. We propose CWIN, a novel window-based drift detection technique that exploits the two-sample location-scale Cucconi test. Preliminary results show that CWIN surpasses the state-of-the-art KSWIN in 10 of 12 data streams tested and is better at successfully detecting concept drifts.