WP-Rank: Rank Aggregation based Collaborative Filtering Method in Recommender System
Sri Lestari, Teguh Bharata Adji, Adhistya Erna Permanasari · International Journal of Engineering & Technology · 2018
Collaborative filtering with a traditional rating-based approach uses interaction records between users and systems to measure similarity, prediction, and generate recommendations. However, traditional rating based cannot capture user preferences of different products. To overcome this issue, the ranking based approach, such as the Borda method has been used. The method takes advantage of rating data to determine the position of the product in the list of preferences as the basis for determining product points. However, the list of the preferences which is merely based on rating data, results in an insufficient accuracy of the recommendation. This paper, therefore, proposed a novel approach namely the WP-Rank aggregation method which maximizes the use of rating data to generate product weight. The experimental results show that the WP-Rank method was superior to the Borda method with NDCG average value difference of 0.0220. However, the WP-Rank method required longer running time with 0.0206 seconds lag from the Borda method.