摘要
针对传统推荐算法所面临的数据稀疏以及在相似度度量过程中对项目属性考虑欠缺的问题,提出了基于项目属性偏好挖掘的协同过滤推荐算法。首先,鉴于项目属性的多标签性质,在项目属性偏好挖掘过程中提出了对多标签属性的概率稀释的处理方法,挖掘用户项目属性偏好;然后,提出了双阈值相似度计算方法计算用户间的相似度,求得基于属性偏好的近邻集合;最后,利用用户-项目评分矩阵,采用基于用户的协同过滤算法对目标用户进行评分预测和推荐。实验结果表明,提出的算法不仅将项目属性因素融合到推荐算法中,而且有效地缓解了数据稀疏问题,同时其在推荐精度上也有不小的提升。
The traditional User-based Collaborative Filtering( UCF) faces several problems of sparse user ratings and lacks consideration of the item attributes. This paper proposed an algorithm named item-attribute-based collaborative filtering.Firstly, according to the multi label property of items, a method to decompose the probability of the multi-tag attribute in the process of mining the preference of item attributes was proposed for mining the user' s preference on item attributes; secondly,the dual-threshold similarity computing method was used to compute the similarities between users and the neighbors based on attribute preference; thirdly, the user-item rating matrix and UCF method were used to predict the rating of some items and recommend items for the objective user. Experimental results show that the algorithm proposed in this paper not only integrates the factor of item attributes into recommendation algorithm, but also alleviates the data sparse problem effectively, and at the same time it outperforms other recommendation algorithms in the aspect of recommend precision.
出处
《计算机应用》
CSCD
北大核心
2017年第A01期262-265,共4页
journal of Computer Applications
基金
北京高等学校教育教学改革面上项目(2013-ms041)
关键词
多标签
偏好挖掘
相似度计算
协同过滤
推荐算法
multi-tag
preference mining
similarity computing
collaborative filtering
recommendation algorithm