A Lightweight Network Architecture for Enhanced Music Recommendation

Music recommendation systems (MRS) have improved the listening experience for many users by offering personalized suggestions, enhancing their mood, and increasing productivity. Machine learning (ML) techniques are powerful tools in deciphering user preferences, and MRS were able to offer users personally tailored suggestions. MRS needs to be installed on handy devices like mobiles thus the recommendation models need to be lightweight. This research introduces a new lightweight neural network architecture, the 3-Layer Dense Neural Network (3DLNN), which is optimized for efficient and precise music recommendations. For evaluation, its effectiveness was compared with traditional techniques such as Content-Based Filtering (CBF), Collaborative Filtering (CF), Context Filtering, and Hybrid Systems. Using the Free Music Archive (FMA) dataset, a database containing over 100,000 audio tracks with extensive metadata, the results show the 3DLNN's superiority in terms of precision (0.8), F1-score (0.889), and RMSE (6.462). The model was superior to the others in ranking accuracy and error minimization and outperformed all other methods, while CBF had the lowest performance, with a precision of 0.4 and F1-score of 0.571. Findings indicate that 3DLNN is a strong and scalable framework for music recommendation systems for the next generation and enable new ways toward adaptive neural network-based personalization.