Personalized Korean Drama Series Recommender System Using Singular Value Decomposition

Korean drama series recommender systems aim to assist users in finding Korean drama series that match their preferences. Based on previous studies, there are several recommender systems on Korean drama series with various methods, but many of them are still limited to certain criteria, such as users’ favorite actors or genres, thus reducing the variety of recommendations. Therefore, we propose a recommender system for Korean drama series focusing on the MyDramaList dataset using the Singular Value Decomposition (SVD) method that utilizes other users’ ratings, thus enabling more in-depth identification of user preference patterns and items. The SVD method improves recommendation accuracy by extracting important features and overcoming sparsity issues, resulting in more relevant and personalized recommendations. Based on the test results, the SVD method performed best compared to the baseline K-Means and K-NN methods with an RMSE value of 1.443 and MAE of 0.900, reflecting a high level of prediction accuracy.