针对Web 2.0环境下大众分类系统中用户、资源和标签之间的三元相关关系,本文提出一种基于三部图的用户兴趣扩散模型,据此为用户进行网络资源的推荐.其主要思想是:通过迭代的扩散机制,使目标用户对信息的兴趣依三部图结构扩散至其他的用户、标签和资源上,然后以资源兴趣度排序为依据,在目标用户未曾收藏的资源中产生推荐.该推荐方法的优势在于扩大了推荐范围,避免了数据稀疏对推荐造成的干扰.利用公共数据集进行的实验表明,本文提出的推荐方法其准确率和召回率优于基于二部图用户兴趣扩散的资源推荐结果.%Considering the user-resource-tag ternary relation in folksonomies of Web 2.0,we propose a user tastediffusion model based on tripartite graph to recommend network resources for users. In this model, we introduce the iterative mechanism to the user taste diffusion model and make the target user' s taste on information diffuse to other users, resources and tags in accordance with the tripartite graph structure. Then we recommend the resources which haven't been collected by the given user according to his or her taste degree on them. Our proposed method expands the recommending range and alleviates the disturbance caused by the data sparsity to some extent. We test our method on the public data set and compare it with the user taste diffusion methods based on bipartite graph. The experiment shows our approach performs best.
展开▼
