Concept drift in data streams can seriously affect the performance and stability of data stream classification algorithms and reduce the generalization performance of integrated learning models. To address the Concept drift problem in dichotomous data streams, this paper proposes a modeling method for enhancing inter-base learner diversity based on evolutionary computation techniques. The method first groups each base learner according to its performance on the sliding window. Secondly, the base learning periods are evolved based on evolutionary techniques. Further, the concept of gene flow is introduced to increase the diversity among base learners and improve the prediction performance of Concept drift. The results on real and artificial datasets show that the comprehensive performance of the proposed method is better than other similar methods.