摘要： 目前的用户概貌攻击检测算法无法避免垃圾用户和真实用户的误判现象，从而影响个性化协同推荐系统的精度。为解决该问题，将时间集中性的概念引入到攻击检测中，提出一种基于正态云模型和时间集中性的可疑评分度量方法，并在此基础上给出一种基于攻击检测的用户可信度计算方法。实验结果表明，该方法能够根据用户评分的真实程度为每个用户计算出评分可信度，提高推荐精度。

关键词: 攻击检测, 云模型, 时间集中性, 可信度

Abstract: Recent attack detection algorithms can not avoid misjudgment phenomenon of the spam users and the real users, which seriously affect the accuracy of the personalized collaborative recommendation system. In order to solve the problem, the concept of time concentrated characteristic is introduced into the attack detection. An approach to measure the suspicious ratings based on normal cloud model and time concentrated characteristic is proposed. On the basis of these, an users credibility calculation method based on attack detection is presented. Experimental results show that this method can not only calculate the trust for every user based on the facticity of users ratings but also improve the accuracy of the recommendation algorithm.

Key words: attack detection, cloud model, time concentrated characteristic, credibility

中图分类号: 

• TP309