Emotion-Aware Personalized Music Recommendation System Based on Improved Deep Convolutional Neural Networks

Now a days, the Music recommendation systems aim to enhance user listening experiences by suggesting relevant songs based on user preferences. However, existing systems face challenges in capturing user emotions effectively due to inadequate feature extraction. To overcome these issues, in this research, an Improved Deep Convolutional Neural Networks (Improved DCNN) method is proposed for Emotion-Aware Personalized Music Recommendation System (EPMRS). Initially, the input data is collected from the Million Songs Dataset (MSD) which contains user data, and audio feature, then preprocessing is performed using Label Encoding to convert categorical features into numerical values and Min-Max scaling is used to scale numerical values. Further Improved DCNN and Weighted Feature Extraction (WFE) are employed for feature extraction whereas Improved DCNN extracts latent features and WFE derives implicit user ratings to capture the relationship between user data and music data. After extracting features, EPMRS recommends music to users by analyzing their implicit ratings and matching them with songs that share similar emotional and acoustic characteristics. The proposed Improved DCNN model archived better results in terms of accuracy with 88.90% compared to existing DCNN.