协同推荐系统检测模型的一种优化方法

An Optimal Method for Detection Model in Collaborative Recommender Systems

研究表明协同推荐技术容易受到攻击。由于现有的检测模型对低填充规模攻击的检测效果不理想,本文结合检测模型特点,改进Pearson相似度计算方法。其思想是,降低共同评分项目对用户相似度程度的影响,从而降低填充规模较小的攻击数据与真实用户之间的相似度。实验结果表明该方法对低填充规模攻击有较好的抗攻击性。

Research has shown that collaborative technologies recommended vulnerable to attack. While some attack profiles may go undetected at low filler sizes by existence detecting models. In this paper,considered of detection models,we improve the method of Pearson similarity computation method. This method is able to reducing the common evaluation item＇s impact to compute the user＇s similarity,and as a result,it will decrease the similarity between the attack users and the real users. The experimental results show that it is effective to raise the systems robust against attack with low filler sizes.