基于信任模型的社区发现与协同过滤推荐研究与应用

Research and application of community discovery and collaborative filtering recommendation based on trust model

文章改进NMF社区发现算法(Nonnegative Matrix Factorization,NMF),得到基于双属性矩阵的NMF社区发现算法(Double Attribute Matrix Nonnegative Matrix Factorization,DAMNMF),在其社团内部进行推荐。因为数据集评估不足可能造成稀疏性问题,使得推荐的效果变差。针对上述问题,将社区发现和信任模型相结合,得到基于社区发现内部信任模型协同过滤推荐,有可能将那些试图影响推荐准确性的恶意用户去除。考虑到信任可以缓解这些问题,则在社区发现中加入信任这个概念,即基于信任模型双属性矩阵非负矩阵分解(Trust Model Double Attribute Matrix Nonnegative Matrix Factorization,TMDAMNMF)社区发现与协同过滤推荐,在真实数据集上进行实验,研究结果表明推荐效果得到了进一步的提升。

In this paper, the NMF community discovery algorithm(Nonnegative Matrix Factorization, NMF) is improved. The NMF community discovery algorithm(Double Attribute Matrix Nonnegative Matrix Factorization,DAMNMF) based on Double Attribute Matrix is obtained and recommended in its community. Sparsity is a problem due to inadequate data set evaluation. To solve the above problems, community discovery and trust model are combined to obtain collaborative filtering recommendations based on community discovery internal trust model, which may remove malicious users who try to affect the accuracy of recommendations. Considering that trust can alleviate these problems, the concept of(trust is added into community discovery, namely Trust Model Double Attribute Matrix Nonnegative Matrix Factorization, TMDAMNMF) community discovery and collaborative filtering recommendation,and experiments on real data sets show that the recommendation effect has been further improved.