基于近邻评分预测的协同过滤推荐算法

摘要

With the number of users growing and types of services provided by websites increasing, these sites are faced with the problem of how to provide more precise recommend things that their users might be interested. Traditional User-Item Ratings matrix calculation method of similarity between items is not accurate enough. When the User-Item rating matrix is sparse, it is not accurate or even impossible to deal with. This similarity ratings in the project considered the time information;in the calculation of the project similarity ratings it combines rating similarities and project properties characteristic similarities. Experimental results show that the algorithm is better than the traditional method.It can deal with the data sparseness problem and improve the accuracy of the recommended results.%随着用户数量和网站提供的服务种类的不断增加，这些网站都面临着怎样更精准的给自己的用户推荐他们可能感兴趣的东西。传统的在用户-项目评分矩阵上计算项目之间相似性的方法不够精确，而且当用户-项目评分矩阵很稀疏的时候误差很大甚至无法处理。文中在项目评分相似性计算中考虑了时间信息，在计算项目相似性中融合了项目评分相似性和经过加权处理的项目属性特征相似性。实验结果表明，该算法较之传统的方法能够较好的应对数据稀疏问题，同时提高了推荐结果的精确度。