众筹项目的个性化推荐:面向稀疏数据的二分图模型

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.

Personalized recommendation of crowd-funding campaigns:A bipartite graph approach for sparse data

Bipartite graph is a global optimal algorithm, which enables direct recommendation of crowd-funding campaigns. In our method, PersonalRank is applied to calculate global similarity for a network in an iterative manner. It can be applied to recommendations where Cosine similarity function is ineffective. Furthermore, we propose a bipartite graph based collaborative filtering (CF) by combining CF and PersonalRank. The nodes are classified into one of the following two types: user nodes and item nodes. For any two types of nodes, the new model calculates the global similarity between the nodes by PersonalRank, and obtains the recommendation list through CF algorithm. Experiment results show that the bipartite graph based CF achieves better performance for the extremely sparse data from crowd-funding community.