Enhancing Music Course Learning Efficiency Using a GMF-MLP-Based Collaborative Filtering Recommendation Algorithm in Big Data Environments

Aiming at the problem of mismatch between music courses and students’ levels in big data environments, this study applies a collaborative filtering recommendation algorithm that integrates Generalized Matrix Factorization (GMF) and Multilayer Perceptron (MLP) to improve course learning efficiency and academic performance. Independent embedding vectors are generated for each user and item. In the GMF part, the inner product of the user and item embedding vectors is calculated to capture their collaborative relationship. In the MLP part, the embedding vectors of users and items are concatenated into a long vector and passed as input to multiple fully connected layer networks. Leaky Rectified Linear Unit (ReLU) is added after each layer for nonlinear activation. The nonlinear relationship between users and items is captured by converting the input layer by layer, and the Dropout layer is added to reduce overfitting and accelerate convergence. The outputs of the GMF and MLP algorithms are integrated, and the weighting coefficients are adjusted using an additional fully connected layer to adjust the model contribution. The Layer-wise Adaptive Moments optimizer for Batch training (Lamb) optimizer is used for distributed training optimization, and the final score prediction is formed through the Tanh function. The research results show that the recommended matching Mean Absolute Error (MAE) value of the GMF-MLP algorithm is 0.035; the Root Mean Square Error (RMSE) value is 0.048; the learning task error rate in the 10th month of the test is about 3.8% lower than that of ordinary course teaching; the academic performance in the 12th month is 25.8% higher than that of the first month. The method used can significantly improve students’ academic performance and optimize the learning efficiency of music courses.