Abstract: In the process of the rapid development of mobile networks, music recommendation systems (MRSs) have experienced considerable success in recent years. Conventional music recommendation systems are, however, in general based on the simple user–track relationships or the content of songs and recommend songs according to intrinsic factors. Furthermore, they do not consider the users’ contextual factors towards providing them with a more interpretable, efficient and smart recommendation experience. To address these issues, we propose anovel Heterogeneous Information Network-based Music Recommendation System (HIN-MRS). By considering the extrinsic factors, such as contextual factors, internal factors, such as the user’s personalized preference, and the heterogeneous relationship between items of song information, this method can perceive the user’s music selection from multiple aspects, automatically maintain the user’s playlist and improve the user’s music experience. First we used the obtained textual data to extract the user’s music preference to provide the topic which is usually related to the contextual factors, by means of which an HIN-MRS can realize the perception of the mobile environment. Second, after determining the topics, we built a small-scale HIN of songs according to topics and used a graph-based algorithm to generate recommendations. The recommendation method based on an HIN renders the recommendation process more efficient and the recommendation results more accurate and increases the user’s satisfaction. The results of our final experiments also prove the significant advantages of the proposed model over the conventional approaches.

Keywords: MRS, HIN, SVD, PMF, LDA, SDA, CNN, LRU

| DOI: 10.17148/IARJSET.2023.10754