Music Personalized Recommendation System Based on Deep Learning

In the Internet age, these have brought great convenience for users to obtain personal preference information. Among them, personalized recommendation algorithm is a very key problem. The music personalized recommendation system studied in this paper combines DL (Deep learning) technology, uses CNN (Convolutional Neural Network) to predict the hidden features of music, and obtains the low-dimensional vector representation of music features. Combined with the hidden representation of user preferences, it learns the hidden layer output of auxiliary information through AE (Autoencoder) and then integrates it into the traditional personalized recommendation algorithm. Finally, a reasonable personalized recommendation is generated for relevant users. According to the experimental results, with the increase in the number of prediction scores, the value of MAE (Mean Absolute Error) decreases continuously. The MAE of this model is lower than other models, and it has a better recommendation effect. The reliability of the personalized music recommendation system established in this paper is verified.