论文部分内容阅读
为了有效地解决协同过滤算法中新项目难以推荐的问题,文中提出了一种对项目矩阵进行划分的方法。其基本思想是,首先利用分类树算法划分项目矩阵并计算项目间的相似度,在此基础上缩小近邻搜索的范围和需要预测的资源数目。通过用户对已有项目的评分排列顺序和项目间相似性预测用户对新项目的评分。实验结果表明:基于项目矩阵划分的协同过滤算法有效地解决新项目推荐困难的问题,显示出了比传统推荐算法更好的推荐质量和扩展性。
In order to effectively solve the problem that the new item in the collaborative filtering algorithm is difficult to recommend, a method of partitioning the project matrix is proposed in this paper. The basic idea is to use the classification tree algorithm to divide the project matrix and calculate the similarity between items, and then narrow the scope of the nearest neighbor search and the number of resources to be predicted. The user’s rating of the new item is predicted by the user’s ranking of the existing items and the similarity between the items. The experimental results show that the collaborative filtering algorithm based on project matrix partition effectively solves the problem of new project recommendation difficulty and shows better recommendation quality and expansibility than the traditional recommendation algorithm.