Personalized Recommendation Method for Course Ideological and Political Teaching Resources Based on Data Mining

In order to improve the personalized recommendation effect of ideological and political teaching resources in courses, a multi-layer microservice architecture recommendation system was constructed based on data mining technology. The system achieves multimodal data collection through distributed crawlers, extracts semantic features using BERT BiLSTM model, and combines dynamic construction of knowledge graph and user profile to achieve precise resource matching. Adopting a dual layer recommendation model, the improved collaborative filtering algorithm is integrated with the Wide&Deep model to enhance recommendation accuracy. The experimental results show that the system outperforms traditional methods in multiple performance indicators, especially in cold start scenarios where the effect is significant. Research shows that through multi-dimensional feature fusion and dynamic interest capture, the system can effectively improve user learning time and resource click through rate, providing an intelligent solution for recommending ideological and political teaching resources.