Discovery of hidden similarity on collaborative filtering to overcome sparsity problem

초록

This paper presents a method for overcoming sparsity problem of collaborative filtering system. The proposed method is based on an intuition on a network of human (such as a friendship network with friends of a friend). This method increases the density of similarity matrix and the coverage of predictions. We use sparse training data to test the sparsity of real-world situation. Consequently, experimental results show that this method increases coverage and f-measure especially for sparse training data.