对用户提供个性化的服务推荐是铁路信息服务系统提高服务质量的手段之一，传统的协同过滤推荐算法的应用过程中，当用户和项目数目较大时，用户-项目矩阵极度稀疏且用户评价向量的维度不同，传统的相似性比较方法无法很好地处理此类问题，降低了推荐算法的推荐质量。本文提出基于云模型的相似性度量方法，使用云模型雾化特征对逆向云算法进行补充，对于给定的数据向量，可以将其转换成云，使用云模型数字特征进行数据表示。云模型的期望和熵决定数据对应概念的内涵相似度，熵和超熵可以反映其外延相似度，比较云之间的相似度可得到数据本身的相似程度。应用MovieLens标准测试数据集，与传统的相似性度量方法比较，实验结果表明，基于云模型相似性度量的协同过滤推荐算法推荐质量高，可为铁路信息化服务推荐提供技术积累与指导。%Personalized service recommendation was one of important directions of China railway information service system. In collaborative ifltering recommendation algorithm, the traditional similarity measurement methods could not deal with large number of users and items which formed a sparse User-Item matrix. This article proposed the similarity measure method based on Cloud Model, applied atomized feature of the Cloud Model to the reverse cloud algorithm. A given data vector could be converted into a cloud. Quantitative data was represented with the numerical characters of the Cloud Model. Cloud similarity was depended on two aspects, such as the concept of connotation similarity and extension similarity. The connotation similarity of data in the corresponding concept was depended on the entropy and expectation of the Cloud Model and extension similarity was reflected by the entropy and excess entropy. A new collaborative filtering recommendation algorithm based on the Cloud Model similarity measurement method was constructed and the experiment result showed that the new algorithm was with reliable and accurate performance.
展开▼
