基于信任和不信任关系的实值受限玻尔兹曼机推荐算法

在社交网络与电子商务快速融合的背景下,将基于信任关系的推荐技术应用于电子商务领域实现个性化推荐已得到广泛研究.现有推荐算法鲜有考虑用户间不信任效应,导致社交信任度量过于保守,较大地影响了推荐系统准确性.针对现有推荐算法忽视不信任关系导致的非对称效应缺陷,本文提出一种结合信任和不信任的实值受限玻尔兹曼机推荐算法(TDA-RBM),首先建立个人受限玻尔兹曼机,进而运用用户社交行为特征信息分析用户信任与不信任关系并进行度量,在此基础上构造信任-不信任监督机制并用于TDA-RBM方法的优化,同时对该方法的有效性进行分析.通过Epinions数据进行的对比实验表明了TDA-RBM方法的有效性以及不信任关系的引入能有效提高推荐准确性.

The real-value restricted Boltzmann machine recommendation algorithm based on trust-distrust relationship

The recommendation technology based on trust relationship is applied in the field of mobile e-commerce, and the use of social trust network to realize personalized recommendation has been widely studied. The inaccurate measurement of social trust has great influence on the accuracy of recommendation system. The paper is concerned with the problem that trust mechanism is ignored the asymmetric effect brought by the distrust relationship, and then restricts the accurate measurement of social trust. A novel algorithm of recommendation is proposed which is a real-value constrained Boltzmann machine recommendation algorithm based on trust-distrust relationship (TDA-RBM), establish restricted Boltzmann machine for each user, using normalized real value analysis and training process of RBM. Construct trust-distrust supervision mechanism, trust, distrust and the trust transfer are introduced, and the trust as weights for the TDA-RBM prediction score. Finally, we experimented on Epinions datasets and the comparison experiment proves that the TDA-RBM algorithm can effectively improve the accuracy of recommendation, which shows the effectiveness and excellency of the algorithm.