融合正态分布函数相似度的协同过滤算法

传统协同过滤推荐算法的相似度量方法仅考虑用户间共同评分，忽略了用户间潜在共同评分项等信息量对推荐结果的影响。针对上述问题，设计了一种正态分布函数相似度量模型，此模型考虑了用户间的共同评分、共同评分项目数、以及用户的评分值，据此提出了融合正态分布函数相似度的协同过滤算法，该算法通过综合多种评分因素利用正态分布函数和修正的余弦相似度共同度量用户间的相似关系。实验结果表明，在两种数据集上与几种不同的推荐算法相比，该算法的相似度量方法提高了目标用户查找邻近用户集合的准确率，提高了系统的推荐质量。

Collaborative filtering algorithm fusing similarity of normal distribution function

The similarity measure method of traditional collaborative filtering recommendation algorithm only considers the common rating among users, ignored the effect of information such as potential common rating items among users on the recommendation results. Aiming at addressing the above problems, this paper designed a similarity metric model of normal distribution function. This model took into account the common rating between users, the number of common rating items and the user''s rating value, based on which a novel collaborative filtering algorithm fusing the similarity of normal distribution function is proposed. The proposed algorithm utilized the normal distribution function and the adjusted cosine similarity to jointly measure the similarity between users by integrating multiple scoring factors. Experimental results on two data sets show that, compared to several different recommendation algorithms, the similarity measure method of the proposed algorithm improves the accuracy of searching adjacent user sets through target users and the quality of the recommendation system.