With the rapid development of the Internet and information technology, the availability of music resources has experienced significant growth. However, along with this abundance, users often find it difficult to navigate through such an extensive collection and locate the specific songs they are looking for, leading to a considerable investment of time and effort. Thankfully, the emergence of music recommendation systems has addressed this problem effectively. These systems leverage advanced algorithms to swiftly help users discover music that aligns with their preferences, thereby saving their valuable time and energy, while also contributing to the economic success of the music platforms. This research centers around the recommendation of music using Transformer-based frameworks. To achieve this, we harness the power of the PyTorch framework to build a comprehensive network model that takes into account essential factors, such as music information, user profiles, contextual details, and historical user behavior data. An efficient transformer module is derived and serves as the backbone of the network model, facilitating the generation of a top-k recommendation list tailored to each user’s preferences. The module combines the attention free transformer and the convolutional layers. In evaluating our approach, we rely on established metrics, such as accuracy and recall, to assess the level of user interest. Experiments affirm the superiority of our approach, surpassing the performance of existing methods in the field.
Xu et al. (Sun,) studied this question.