The increasing maturity of data analysis technology provides technical support for personalized recommendation of students’ learning path. This paper establishes a multi-dimensional student portrait labeling system by collecting behavioral data and ability characteristics of vocational education information technology students. Aiming at the limitations of the traditional single-view clustering method, a multi-view deep clustering model is selected to integrate the students’ outcome and process characteristics, explore the complementarity of different view data, and improve the accuracy of student clustering. Combined with the dynamic generative recommendation strategy, differentiated learning resource sequences are matched for different categories of students to achieve learning path optimization. The model is applied to real vocational education information technology majors to verify the personalized learning assistance effect of the model. The results show that students can be clustered into 4 categories according to 9 categories of information technology ability characteristic levels. This paper’s model scores more than 0.75 on five performance indicators, which is better than the comparison model. In the control experiment, the experimental group using this paper’s model to assist learning scored more than 60 points in each characteristic competency, and the rate of strong agreement in student satisfaction in the experimental group was more than 70%.
