A music recommender system based on compact convolutional transformers

In recent years, music streaming services have garnered a surge in popularity. Maintaining a continuous flow of music that aligns with user preferences, commonly known as the continuity problem, has become a significant issue in this domain. To address this challenge, developing Music Recommender Systems (MRSs) that can automatically search through vast music libraries and suggest appropriate songs to listeners is crucial. To this end, this paper proposes a Compact Convolutional Transformer (CCT) model for improving the feature selection process and thus addressing the continuity problem based on music genres. The model extracts latent features from Mel-spectrograms generated from raw audio songs. Then, the cosine similarity measure determines the similarity between feature maps to recommend the most relevant songs. Several methodologies, including two state-of-the-art CRNN models, are used to benchmark the model's performance. The experimental results demonstrate that the proposed model significantly outperforms the current state-of-the-art models in terms of precision, recall, F1 score, and overall accuracy while having significantly fewer parameters.