Book Recommender System Using Singular Value Decomposition Combined with Slope One Algorithm

As information spreads on the internet, users need more personalized information. A recommender system can be applied to achieve this. Singular Value Decomposition (SVD) is a method that can be applied to a recommender system to find hidden features in the relationship matrix between items and between users. However, the recommender system faces several problems, one of which is data sparsity. The SVD algorithm itself considers an empty rating value as 0. In solving this problem, the SVD algorithm can be combined with rating predictions using the Slope One algorithm. The Slope One algorithm is applied to complete the empty rating data so that the model can be trained with more complete data. So, the Slope-SVD method can improve the accuracy of the recommender system. This study aims to apply the Slope-SVD algorithm to a book recommender system using a rating dataset from the Goodreads web. The performance of the system was tested with the Mean Square Error (MAE). The test results from this study indicate that combining the Slope One algorithm with SVD gives better results than each Slope One or SVD alone.