摘要
为了解决用户评分数据稀疏性问题和传统相似性计算方法因严格匹配对象属性而产生的弊端,结合项目分类和云模型提出了一种改进的协同过滤推荐算法。首先,按项目分类得到类别矩阵;然后利用云模型计算类内项目间的相似度并获取具有最高相似度的邻居项目的评分,为类内未评分项目进行预测填充;再利用云模型计算类内用户间的相似度得到用户邻居,最后给出最终的预测评分并产生推荐。实验结果表明,该算法不仅有效地解决了数据稀疏性及传统相似性方法存在的弊端,还提高了用户兴趣及最近邻寻找的准确性;同时,该算法只需计算新增用户或项目所在的类别即可,大大增强了系统的可扩展性。
In order to solve problem of data sparseness in user rating matrix and the drawback of attributes' strictly matching in traditional similarity calculation method, this paper presented an improved collaborative filtering recommendation algorithm by combining the item classification and cloud model. This method firstly utilized the item classification information and cloud model to compute items inner-similarity, and then got the scores from neighbor items which had gotten the highest similarity and used their scores to forecast the unrated inner-class items. Secondly, this method obtained the user' s neighbors through inner-class user similarity gained from the cloud model computing, then gave the final forecast grade and carried out the recom- mendation. Experimental results show that this algorithm is not only an effective solution to data sparseness and the drawbacks of traditional similarity method, but also improves the accuracy of user interest and nearest neighbor search. At the same time, the algorithm that only calculates the categories which adds the new users or items, it greatly increases the scalability of the system.
出处
《计算机应用研究》
CSCD
北大核心
2012年第10期3660-3664,共5页
Application Research of Computers
关键词
云模型
项目分类
协同过滤
项目相似性
推荐系统
cloud mode
item classification
Collaborative filtering
item similhrity
recommendation systems