A New Interval Preference Model and Corresponding Fuzzy Similarity Measure for Collaborative Filtering
Yong Wang, Pengyu Wang, Xuhui Zhao, Zhuo Liu, Leo Yu Zhang · 2020
Collaborative filtering as one of most popular recommendation engines generates high-quality candidate item sets for users, which reduces their heavy burden on selecting desired products or services in an ocean of information. The basic idea of collaborative filtering is to extract preferences of users according to their historical behaviors, such as ratings. Rating information is able to easily collected and reflects the opinions of users on different items to some extend. However, previous publications indicate that different users have different rating tendencies, thus the same ratings may represent different preferences. To address the issue, we extend the previous preference model and present a new interval preference model. Based on the model, we devised the corresponding fuzzy similarity measure for collaborative filtering. The experimental results indicate that the proposed method performs well on Movielens dataset and has potential application value.