Improved Ranking Based Collaborative Filtering Using SVD and Borda Algorithm
Muhammad Iqbal Ardiansyah, Teguh Bharata Adji, Noor Akhmad Setiawan · 2019
Ranking based collaborative filtering (CF) as opposed to the well-known memory based collaborative filtering has better performance in accommodating user preferences. However, ranking based collaborative filtering still suffers both sparsity and scalability issues. This work incorporates the Singular Value Decomposition (SVD) technique to reduce the user-item rating matrix dimension to solve the scalability and sparsity of many unrated items faced by ranking based collaborative filtering. From the evaluation, the proposed method acquires 4 times higher coverage but 33.3% lower F1-score compared to BordaRank method. The proposed method also gets 6 times higher F1-score and 17.8% higher coverage compared to pure SVD method. The running time of the proposed method shows that on lower neighborhood size, the proposed method gets the fastest and the best performance. Thus, incorporation of SVD to ranking based CF can be a good approach to overcome sparsity and scalability problems.