Abstract: Collaborative filtering has been successfully applied to various recommendation systems. Unfortunately, with the tremendous growth in the amount of items and users, the lack of original rating poses some key challenges for recommendation quality. To address this problem, the paper explored a new hybrid CF approach which improved the traditional similarity coefficient computation combining the portal website natural structure, then the missing preference values were predicted with new similarity based on the item-based CF. The improvements made an increasing intercross of the rating matrix for alleviating sparsity. The experimental results show the proposed algorithm outperforms the traditional CF, and it can recommend potential preference pages for visitors.

摘要： 目前协同过滤技术已经被成功运用到各种推荐系统中,但是随着资源种类的不断膨胀与用户日益的增加,用来评判的数据矩阵越来越稀疏,严重影响了推荐质量。为此设计了一种混合新算法,对基于项目的协同过滤算法提出两个改进方法：首先根据网站的层次结构信息改进了传统的相似度计算方法；其次增加了预测缺失兴趣值的算法,使用户的交叉兴趣点增多,有效缓解了稀疏性的问题。实验结果证明了新算法具有较高的推荐精度,能够找到用户潜在的兴趣页面。

关键词: 推荐系统, 协同过滤, 数据预测, 数据稀疏性, recommendation system, collaborative filtering, data prediction, data sparsity