Music recommendation method based on time residence and state residence mixed model

Abstract

The invention discloses a music recommendation method based on a time residence and state residence mixed model, which comprises the following steps: expanding five elements of a traditional hidden Markov model into seven elements, calculating and predicting music which is possibly interested by a user at the next moment in behavior information of interaction between the user and the music, and forming a user preference list. A user preference vector is extracted from a user score matrix by using an SVD matrix decomposition algorithm, meanwhile, potential features of song audio signals are mined by using a convolutional neural network, and song tag features are extracted by combining an embed layer with a long short-term memory artificial neural network LSTM, the obtained music potential feature vector and the user preference vector are trained, and the two are trained at the same time to obtain a candidate list; and finally, the two lists are reordered by utilizing a rating function to obtain a diversified optimal recommendation list. According to the method, the recommendation accuracy is improved, information overload is effectively solved, meanwhile, the problems of cold start and the like in a recommendation system are relieved, and the recommendation diversity and robustness are improved.

Description

technical field [0001] The invention belongs to the technical field of music recommendation, in particular to a music recommendation method based on a time-resident and state-resident hybrid model. Background technique [0002] The main task of the traditional recommendation method is to improve the accuracy of the recommendation, but ignores the consideration of the diversity and robustness of the recommendation effect, and most of the research and expansion of the music recommendation system focus on improving the accuracy of the recommendation, solving User data sparsity and cold start problem. The accuracy of the recommendation system is to judge whether the prediction of a single piece of music is accurate. As time goes by, users may lose interest in the recommended music list, and the diversification of recommendations can improve user satisfaction with the recommended list. In recent years, there have also been many articles on recommendation diversity research. For ...

AI Technical Summary

Problems solved by technology

However, in recent years, there has been no recommended system that has been studied, which can not only improve the diversity of recommendations, but also improve the robustness of the recommendation system.

Images