Algorithm Design and Empirical Analysis of Learning Path Recommendation System for Financial Engineering Course

In view of the increasing demand for talents caused by the complexity of financial market and the diversity of financial products, the traditional teaching mode is insufficient. Therefore, this paper proposes a hybrid recommendation algorithm combining collaborative filtering (CF) and deep learning (DL), and customizes personalized learning paths for students by analyzing students' historical behaviors, interest preferences and potential relationships among courses. The system architecture covers four modules: front-end display, data processing, algorithm core and user interaction, and the interaction between modules is realized through data interface. In the implementation of the algorithm, firstly, the user-course matrix and feature vector are constructed, and the similarity is calculated by CF and the nearest neighbors are searched. At the same time, the relationship between students and courses is learned by DL model, and finally the two results are fused by weight to generate a recommendation list. Based on the study records of about 5000 students and the information of more than 300 courses in an online platform in two years, the experiment compares the effects of mixed recommendation algorithm and CF or DL alone. The results show that the hybrid algorithm is superior to the single algorithm in accuracy (0.756), recall (0.682) and F1 score (0.718), which verifies its effectiveness in improving recommendation accuracy, personalization and diversity of learning paths.