中文摘要：

推荐系统是电子商务系统中最重要的技术之一．协同过滤技术是当今应用最普遍的个性化推荐算法．针对用户评分数据的极端稀疏性和算法的可扩展性，首先利用云模型计算项目间相似度来预测用户对未评分项目的评分，来增加用户评分数据，再根据项目分类信息将用户．项目评分矩阵转换为用户一类别矩阵，降低了评分矩阵的维度，最后利用云模型计算用户间相似度，得到目标用户的最近邻居．实验结果表明。该方法具有较小的MAE，提高了推荐系统的推荐质量．

英文摘要：

Recommendation system is one of the most important technologies in E-commerce. Collaborative filtering is the most prevalent algorithm of personal recommendation. Aimed at the sparsity of user rating data and algorithm scalability, firstly it predicts the item ratings that users have not rated by calculating the item similarity based on the cloud model, in order to add user rating data. Secondly user-item rating matrix is converted to user-classification rating matrix according to classification information of resource, which reduces the dimension of the rating matrix. Finally it calculates the similarity of users by using to find the target users＇ neighbors. The experimental results shows that the algorithm has smaller MAE value and can improve the recommend quality of the recommender system.