A probabilistic linguistic and dual trust network-based user collaborative filtering model

Recommendation models for network information that are generally based on user ratings fail to utilize user online behaviours such as reviews and likes, which indicate users' opinions, attitudes, and emotions. To sufficiently represent user preferences and further enhance recommendation accuracy, a probabilistic linguistic and dual trust network-based user collaborative filtering (PLDTN-UCF) model is proposed in this paper. To reflect the uncertainty of user ratings, an easy-to-use function is proposed to transform the personalized semantics of online reviews into a probability distribution that corresponds to user ratings and construct a probabilistic linguistic rating matrix. Then, the calculation approach of traditional user ratings-based trust network is improved by integrating probabilities to represent the fuzziness of trust. Furthermore, a dual trust network is constructed to represent multi-source interpersonal trust based the on an online behaviours-based trust network and probabilistic linguistic rating matrix-based trust network. Finally, the proposed model is compared to state-of-the-art models using the Douban movie dataset to assess its performance.