<p>Concept drift refers to changes in the data distribution of streams over time. When multiple correlated data streams experience drift simultaneously, it results in group concept drift, posing significant challenges for detection. This study addresses group concept drift in multi-stream environments by extending traditional single-stream detection approaches. To improve computational efficiency, a mean-based variance estimation strategy is introduced. Leveraging the Bennett inequality, a mean-based Bennett bound is constructed to form a novel drift discrimination mechanism. A multi-stream detection strategy is further proposed to enhance sensitivity to correlated drifts across streams. Based on these designs, we propose a group drift detection method based on Bennett bound (BGDDM). Additionally, an adaptive three-dimensional performance metric is developed to evaluate detection accuracy, time efficiency, and delay. Experimental results on both real and synthetic datasets demonstrate that BGDDM achieves effective and reliable performance in detecting group concept drift.</p>