Context-aware music recommendation method based on graph embedding model

Abstract

The present invention discloses a context-aware music recommendation method based on a graph embedding model. The method comprises: S1. extraction of a music feature based on a graph embedding model; S2. extraction and modeling of a user global music interest and a contextual music interest; and S3. context-aware music recommendation. According to the method disclosed by the present invention, a feature of music is extracted from playback data of a user and metadata of the music by using a graph embedding model; then a global music interest and a contextual music interest of the user are acquired from a full playback record and a recent playback record of the user; and finally, the global interest and the current contextual interest of the user are comprehensively considered during recommendation, so that recommended music can comply with the real-time demand and preference of the user, and therefore, the search cost of the user is reduced and satisfaction of the user is improved.

Description

technical field [0001] The invention belongs to the technical field of data mining and recommendation, and in particular relates to a context-aware music recommendation method based on a graph embedding model. Background technique [0002] With the development of the digital music industry, more and more online digital music providers have emerged, allowing users to listen to favorite music anytime and anywhere. For example, Apple's online music store provides more than 30 million digital music. At the same time, the massive amount of music data increases the difficulty for users to find the music they are interested in. In addition, users' preferences for listening to music usually change with time, space, weather, and physical conditions. Traditional music recommendation systems are no longer suitable for personalized mobile network services. In recent years, context-aware hybrid music recommendation system has become an emerging research field by introducing context info...

AI Technical Summary

Problems solved by technology

[0004] However, the existing technology only considers the user's context information and partial assistance in the matching process of music and users, lacks in-depth analysis of music content, and believes that all music is homogeneous, and different attributes of music come from users in different situations. Different preferences for music, that is, different music is differentiated by the user attributes of music, thus ignoring the context attributes of music itself as a type of multimedia file

This recommendation method is too subjective, reduces the coupling between users and music, and does not combine auxiliary information of music, such as music playback sequence and metadata, which affects the accuracy of the recommendation system

Images