The cosine similarity approach of the collaborative filtering customized recommendation algorithm is used in this work to first compute the trust degree among users. Next, it identifies the closest neighbor set for users as an addition to the similarity degree, then predicts users’ ratings of items through an optimized item prediction rating strategy based on the ratings of users’ nearest neighbor set and recommends the items with higher prediction ratings to users. Then, we examine the goals and standards for talent development under the “1+X” hybrid certificate and modern apprenticeship systems and explore and build a suitable talent training model for higher vocational mechatronics technology majors. Finally, the research sample is selected from the electromechanical majors of higher vocational institutions, and a collaborative filtering, personalized recommendation algorithm is used to analyze the study sample data set’s data. The findings indicate that T=15 is a preferable option to guarantee a specific level of accuracy since the MAE of the collaborative filtering customized recommendation algorithm at T=30 and T=60 is lowered very little compared to that at T=15. This study provides informative opinions for universities to innovate and reform the talent cultivation mode and solve a series of problems in the current talent cultivation mode so that the quality of talent cultivation in Chinese universities can be improved continuously.