A collaborative filtering recommendation method based on entropy similarity and dynamic trust

Abstract

The invention discloses a collaborative filtering recommendation method integrating information entropy similarity and dynamic trust. The method is based on two similarity calculation methods of information entropy similarity and trust implicit similarity of score difference, and constructs a comprehensive similarity calculation model to alleviate the problem that similarity of cold start users isdifficult to calculate. Then, direct, indirect and global trust calculation models are constructed to reduce false recommendation of unreliable users by integrating the reliability of scoring and recommendation. Secondly, a fusion scoring prediction model of similarity and trust is constructed to complete the scoring prediction and personalized project recommendation for the target users. Finally, we evaluate the effectiveness of the recommendation user scoring, and propose a trust reward and punishment strategy to update the trust neighbor set dynamically for the target user, to suppress thenegative impact of users' random false scoring on the recommendation performance. The experimental results show that the method can improve the recommendation accuracy and reliability of the recommendation system, effectively alleviate the data reliability, data sparsity and cold start problems.

Description

technical field [0001] The invention relates to the technical field of personalized recommendation, in particular to a collaborative filtering recommendation method that integrates information entropy similarity and dynamic trust. Background technique [0002] With the rapid development of information technology and social networks, data resources have exploded, and the problem of information overload needs to be solved urgently. Aiming at the practical problem of how to assist users to efficiently filter and personalize useful information from massive data, recommendation systems emerged as the times require. Currently, recommendation systems can be divided into five categories: content-based recommendation, collaborative filtering recommendation, knowledge-based recommendation, social recommendation, and hybrid recommendation. As an effective strategy to solve the problem of information overload, recommender systems are widely used in e-commerce (Amazon.com, Alibaba, etc....

AI Technical Summary

Problems solved by technology

These classic recommendation algorithms provide key algorithm theories for research in this field. However, with the development of social networks and recommendation systems, the above algorithms cannot well meet the personalized recommendation needs of users. The reasons are as follows:

[0013] 1) There are random or malicious false ratings of users in the recommendation system. Most of the existing methods are based on the assumption that us

Images