摘要： 随着电子商务推荐系统中用户和商品数目的增加，用户商品评分数据集的稀疏性会导致协同过滤推荐算法的推荐质量下降。针对该问题，提出一种基于单分类的协同过滤推荐算法。根据目标用户评分商品对应的类别，选择候选最近邻居集，采用单分类预测用户对商品的评分，以减小目标用户与候选最近邻居所形成的数据集稀疏性。实验结果表明，该算法能提高寻找最近邻居的准确性，从而改善协同过滤的推荐质量。

关键词: 推荐系统, 协同过滤, 数据稀疏性, 单分类, 平均绝对偏差

Abstract: With the increasing number of users and goods in E-commerce recommender systems, the data set sparse of user goods rating reduces the quality recommendation of collaborative filtering recommendation algorithm. To solve this problem, this paper proposes a collaborateive filtering recommendation algorithms based on single-class classificatin. It chooses candidate nearest neighbor set which depending on the target user rating goods corresponding to category and uses single-class classification to predict the values of the user rating. It can reduce the sparse of data set which is formed by the target user and the candidate nearest. Experimental results show that the algorithm is able to increase the accuracy of searching nearest neighbor set, resulting in improving recommendation quality of the collaborative filtering.

Key words: recommendation system, collaborative filteringZ, data sparse, single-class classification, Mean Absolute Error(MAE)

中图分类号: 

• TP311