基于用户项目特征分组的隐私保护算法

针对互联网推荐系统中存在严重的隐私保护问题,在传统推荐系统算法的基础上,引入项目属性相似度的概念,并提出了一种具备保护用户隐私功能的新型推荐系统.系统利用用户的历史评价和推荐系统中项目的属性信息,使用不采集用户个人信息的协同过滤推荐算法,计算出用户对未评价项目的评分预测,形成了一种能够保护个人隐私的推荐算法.结果表明,与其他推荐算法相比,本文算法在推荐准确度和用户隐私保护程度上取得一个较好的平衡,具有较高的实用价值.

Privacy protection algorithm based on user project feature grouping

Aiming at the serious privacy protection problem in the recommendation system of internet, the concept of project attribute similarity based on the traditional algorithm of recommender system was introduced, and a new recommendation system with the function of protecting the privacy of users was proposed. With the historical evaluation of users and the attribute information of projects in the recommendation system, the collaborative filtering algorithm without collecting the personal information of users was adopted in the proposed system, the score forecast of projects without the evaluation of users could be calculated, and the recommendation algorithm which could protect the personal privacy was developed. The results show that compared with other recommendation algorithms, the proposed recommendation algorithm achieves better balance between the recommendation accuracy and privacy protection degree of users, and has good practical value.