Distance Metric Analysis in Recommendation System Using Content-Based Filtering Method

Despite the prevalence of music recommendations in the music industry, the rapid evolution of music and the increasing diversity of musical genres present a formidable challenge in discovering songs that match user’s preferences. The content-based filtering method is a common recommendation system utilizing a music feature-based approach to identify user preferences. The K-Nearest Neighbor (KNN) algorithm is utilized to determine the similarity between songs. However, in order to enhance the quality of music recommendations, it is necessary to employ appropriate distance metrics and threshold values. The study collects song data and associated music features from Spotify to experimentally determine the best distance metric and feature selection threshold. The evaluation metrics employed in this study are root mean square error (RMSE) and mean absolute error (MAE). The results show that Cityblock Distance performs best with an average RMSE of 0.2257, and a threshold value of 0.05 achieves optimal performance with an average MAE of 0.124. These findings indicate that choosing suitable distance metrics and implementing feature selection can enhance the accuracy of music recommendation systems. However, limitations in dataset representation warrant further research with larger datasets and additional factors to improve generalization and user experience in music recommendations.