基于混合用户模型的二分图推荐算法

用二分图来实现个性化推荐的算法越来越受到研究者的注意。文中提出混合用户模型下的二分图推荐算法（MNBI）,针对二分图推荐算法中存在的用户多、项目少时命中效率低的情况用混合用户模型进行改进,同时对于推荐中加权的二分图边的权值用用户集的总体的加权和进行改进。该算法基本思想就是在用户很多的情况下,用混合用户模型对用户首先进行一个预处理生成一定数量的用户集,然后用用户集和项目构成用户集-项目的二分图。通过在Movielens数据集中进行测试的实验结果表明,相比NBI算法,MNBI算法推荐的命中效率有一定的提高,同时对于推荐多样性有所提高,并且在数据冷启动情况下效果较好。

Bipartite Graph Recommendation Algorithm Based on Hybrid User Model

With the bipartite graph to achieve personalized recommendation algorithm has received more and more attention of researchers.Present bipartite graph recommendation algorithm based on the hybrid user model（ MNBI）,aiming at cases of the bipartite graph recommendations algorithm in the presence of multiple users,low project,use hybrid user model to improve,at the same time for the bipartite graph recommendations of weighted edge weights for users to have the overall weighted improved.The basic idea of the algorithm is,using hybrid user model to make a pretreatment for user,generating a certain number of user set when the number of the users is huge,and then use the user sets and project set to construct user-figure two bipartite graph.By focusing on the Movielens data to test the experimental results show that,compared with NBI algorithm,MNBI algorithm recommended hit efficiency is improved,at the same time for recommendation diversity increased,and has good effect in the user data cold start conditions.