Nowadays, the widespread use of multimedia teaching equipment has created the possibility of introducing multi-modal experiential teaching into the classroom. In the multi-modal teaching content recommendation system, the information collection module collects user operation information, including user stay on web pages, user video viewing time, pointer stay information, and corresponding time points for each type of information. The recommendation module selects teaching videos with high consistency by comparing them with the standard question sequence of each teaching video, so that the recommended teaching content meets the real needs of users and improves teaching effectiveness. This project combines management and business big data to design and implement corresponding optimization strategies, compare and analyze the recommendation effects before and after optimization, and evaluate the feasibility of the optimization plan. In the optimization strategy data of management teaching, the accuracy rate before content filtering optimization was 0.80, and the accuracy rate after optimization was 0.85. The accuracy of user behavior analysis before optimization was 0.80, and the accuracy after optimization was 0.83. This article helped to improve the accuracy and personalization of teaching content recommendations.