论文部分内容阅读
二分图模型是一种全局优化算法,本文将二分图模型应用于直接推荐众筹项目,使用PersonalRank算法迭代计算网络节点的全局关联度,从而推荐那些基于余弦相似度的协同过滤不能有效推荐的项目,适用性更加广泛.更进一步,提出将二分图模型与协同过滤算法相结合,首先把网络结构划分为二分图,采用二分图算法得到的两类节点(用户节点,项目节点)之间的全局相似度,再结合协同过滤算法,得到基于二分图模型的协同过滤算法.实验表明,在众筹项目推荐中,由于数据极端稀疏,适宜采用二分图模型来进行相似度计算并进行推荐.
The bipartite graph model is a global optimization algorithm. In this paper, the bipartite graph model is applied to directly recommend crowdfunding projects. The global association degree of network nodes is iteratively calculated using the PersonalRank algorithm, so that those collaborative filtering based on cosine similarity can not be effectively recommended , The applicability is more extensive.Furthermore, this paper proposes to combine the bipartite graph model with the collaborative filtering algorithm, first of all, the network structure is divided into two parts, and the two kinds of nodes (user nodes, project nodes) Similarity and combined with collaborative filtering algorithm to get the collaborative filtering algorithm based on bipartite graph model.Experiments show that in the crowdfunding project recommendation, due to the extremely sparse data, the bipartite graph model is suitable for similarity calculation and recommendation.