协同过滤的一种个性化推荐算法研究

Improved personalized recommendation algorithm in collaborative filtering

在分析传统推荐算法不足的基础上,提出一种稀疏矩阵下的个性化改进策略。首先进行一对一的个性化预测,得到虚拟用户评分矩阵,在此基础上再进行综合预测。该方法避免了传统推荐算法中推荐值与用户相似度不密切相关的弊端,提高了协同过滤的预测精度,尤其是在矩阵极端稀疏情况下的预测精度。最后通过实验验证了算法的有效性和优越性。

This paper analyzed the disadvantages of the traditional collaborative filtering and advanced a kind of personalized predictive strategy：one to one prediction to correct this situation to improve the predictive accuracy in sparse matrix. To prove the superiority of the： proposed algorithm, this paper used cosine similarity and pearson similarity to measure the similarity among users and then produced the predictions using traditional collaborative filtering and the new personalized predictive algorithm. The experimental results prove the validity and superiority of the proposed algorithm at last.