A Novel Deep Neural-based Music Recommendation Method considering User and Song Data

Abstract

This Natural language processing, Computer vision, and speech recognition are among the fields in which deep learning outperforms prior approaches. The majority of deep learning-based music recommendation systems analyze consumers' listening history to understand their temporal preferences. The cold start disadvantage isn't addressed, and these solutions don't properly use the music features. Furthermore, the qualities of music and the consumers' temporal preferences are not organically related, resulting in poor music suggestion performance. In this study, the Deep Neural Based Music Recommendation (DNBMR) method is suggested to address these challenges, which takes into account both music attributes and the customer's temporal favorites. The music data was transformed into one vector and then represented to a low-dimensional space using a Deep Neural Network. Long short-term memory (LSTM) neural networks are applied to collect long and short-term favorites from an individual's listening history as well. By utilizing music metadata and discovering new consumers' preferences as soon as they hear music, In the item aspect, DNBMR addresses the cold start issue. In terms of accuracy, recall, MAP, F-measure, AUC, and user coverage, the experimental findings suggest that DNBMR beats two baseline techniques.