基于社交网络中双重好友及用户偏好的协同过滤推荐

摘要

Collaborative filtering recommendation based on social network suffers the following problems:direct friends information is considered only,weaker capability of profile inj ection attack resistance.To address the problem,a collaborative filtering recommendation algorithm based on double friends’relationships and users’preferences (CF-DFP)is proposed.The trusty friend set is selected from direct friends and potential friends in social network by proper familiarity threshold and is considered as K nearest neighbor candidates (KNNC).Meanwhile,weighted similarity combining users’preferences and ratings is used to find KNN in KNNC and predict ratings for users.Related experiment results on Flixster dataset show that the proposed algorithm can not only improve the accuracy of collaborative filtering,but also resists the profile inj ection attack effectively.%针对目前基于社交网络的协同过滤推荐算法只融入直接好友信息且不能有效防御概貌注入攻击等问题，提出一种融合双重好友及用户偏好的协同过滤推荐算法，通过设置合适的熟悉度阈值在社交网络的直接好友、间接好友中选取可信好友用户集作为目标用户K 近邻候选集，在共同评分项目数的基础上，采用用户偏好相似度与评分相似度的加权相似度作为寻找近邻用户的标准，完成目标用户项目评分预测。在数据集 Flixster上的实验结果表明，融合双重好友关系及用户偏好的推荐算法不仅具有较好的推荐准确率，还具有较强的抗概貌注入攻击能力。