A Music Recommendation Algorithm Based on Convolutional Neural Network Optimization

Abstract

With the advancement of digital multimedia technology, music libraries are growing in size and music resources are getting more plentiful, making it difficult for consumers to easily obtain the songs they want in the music world. For this phenomenon, personalized music recommendation systems have emerged. However, although there are many kinds of existing music recommendation systems, the recommendation effects are uneven, and there are some problems. This study offers a music recommendation algorithm based on neural networks, and conducts systematic experiments to create a network model structure based on a conventional convolutional neural network model, as well as compare and choose model training tuning parameters. Finally, the model is trained and evaluated, and the suggested quality assessment indices are the root mean square error, accuracy, recall, and F1 value. The results of the experiments suggest that the recommendation algorithm presented in this work is both feasible and effective. Unlike other traditional music recommendation algorithms, this paper takes full advantage of deep neural networks' powerful advantage of automatically extracting features and obtaining higher-level music feature representations from audio content while incorporating information on users' historical behavior toward music interactions, effectively solving problems like cold start in recommendation systems.