The application of microteaching in English teaching can not only create a good atmosphere for learning English but also deepen the change of teaching mode and promote English teaching to deep development. In this paper, the output-oriented model is combined with microclasses to establish a blended teaching model for English. For students’ online learning behavior, the K-means clustering algorithm is used to classify them, and the GBDT algorithm and integrated learning Stacking algorithm are combined to predict their performance. Taking the online learning behavior data under the blended teaching mode as an example, a practical analysis was carried out in three aspects: the classification of students’ English online learning behavior, the prediction of academic performance, and the satisfaction of the blended teaching mode. The results show that the number of clusters of English online learning behavior is 5, and the comprehensive score of cluster 1 reaches 91.52. When the progress of blended teaching in the English course developed from 10% to 100%, the accuracy of students’ learning performance prediction increased by 20.97 percentage points. Students’ satisfaction with the “output-oriented method + micro-teaching” blended teaching mode reached 86.34% with grade B and above. The blended teaching mode of combining microteaching and English courses in the digital era can improve students’ independent learning ability and enhance their English application ability.