With the booming development of the digital music industry, the contradiction between the massive music resources and users’ personalized needs is becoming increasingly prominent. Traditional music recommendation systems suffer from problems such as inaccurate feature mining, poor recommendation timeliness, and lagging user interest capture. This article focuses on empowering technology with big data intelligent algorithms to conduct research on music feature mining and personalized recommendation system optimization. Firstly, this article reviews the current research status and technical bottlenecks in the field of digital music recommendation, and clarifies the core research issues; Secondly, this article constructs a big data acquisition and preprocessing system that integrates multimodal music features, and proposes a music feature mining algorithm based on the fusion of deep learning and reinforcement learning; Again, this article designs a personalized recommendation model based on user behavior time series modeling and builds a corresponding recommendation system architecture; Finally, the effectiveness of the proposed algorithm and system was verified through multiple comparative experiments. The experimental results show that the random forest algorithm performs the best in feature mining accuracy, reaching 91.5%, and has the lowest feature redundancy, only 9.8%. The computation time is also relatively short, providing effective technical support for the personalized service upgrade of digital music platforms.
Yanhua Zhu (Thu,) studied this question.