Recommender System F or Educational and Professional Growth

This study presents a course recommendation system that uses machine learning algorithms, such as deep learning, to enhance the reliability and individuality of courses. The system uses data from Peking University's Open Research Data Platform, which contains three main tables: Course Information, Student Information, and Course Selection. The data is combined into a single dataset through joining, and the system transforms education level and course type into numerical features. The model building phase involves developing a neural network with embedding layers for students and courses, LSTM, GRU, and MLP layers. An attention mechanism is also incorporated to improve the accuracy of the model. Hyper parameters are tuned with Optuna to find the best combination of hyper parameters for improved performance.The model's evaluation is evaluated using AUC (Area under the curve) and logarithmic loss. The system can recommend courses even to new students without previous course records. The cold-start problem is addressed by integrating Collaborative Filtering (CF) and Content-Based Filtering (CBF), allowing the system to recommend courses even to new students without prior course records.Future improvements include better machine learning algorithms, such as deep reinforcement learning, and real-time student usage. The system's capabilities are expanded to recommend other educational content, such as books, articles, videos, and learning modes, facilitating personalized learning and helping students achieve their targets. The findings suggest future directions for applying advanced machine learning and deep learning methods in educational recommender systems. The system is highly accurate, user-friendly, and has been proven effective in improving education quality.