- 期刊
An Improved Time-Varying Collaborative Filtering Algorithm Based on Global Nearest Neighbor
摘要
Collaborative filtering has widely been applied to collaborative filtering recommendation systems to filter a tremendous amount of information and screen the information that may be of interest to the target users through the evaluation and feedback of everyone. In contrast, traditional collaborative filtering does not consider the impact of project type and time on user interest changes. A modified time-varying collaborative filtering algorithm (TVCFA) is presented to address the problem of information overload. It is based on a content recommendation algorithm and reduces the static score error of different user features of the similarity expression. It uses the weight values of time and periods to describe the dynamic characteristics of the user-item score. The algorithm's feasibility is verified by simulation using the public data set. It can improve its recommendation model by 6.13% and 2.69%, respectively, compared with the UCF and ICF models in the global nearest neighbor. The experimental results show that the improved dynamic, collaborative filtering algorithm can improve the accuracy and track the dynamic characteristics of users.