The study firstly designed an English education management strategy integrating information technology from three aspects: academic early warning mechanism, early warning students’ intervention and helping measures. After that, the PSO-XGBoost English academic early warning model was constructed based on data mining technology. Learning behavior data from 295 students majoring in English at college A was used to characterize the features. The students were finally categorized into three types using K-means clustering algorithm. Subsequently, the performance test of the PSOXGBoost model proposed in this paper was conducted, and the model of this paper kept each evaluation index above 0.9 when comparing the evaluation indexes with different students’ academic warning levels of 0, 1, and 2. After one and a half years of trial operation of the academic early warning system in English majors of university A, the number of early warning and the number of students in difficulty decreased significantly, and the learning effect of students improved significantly.