多主题受限玻尔兹曼机的长尾分布推荐研究

Research on the Long Tail Distribution Recommendation of the Multi-topic and RBM

随着电子商务的快速发展,网络销售已成为一个重要的商品销售方式,而在线商品销售的长尾效应,也成为电子商务研究中亟待解决的问题.由于对冷门商品的评价数量少,导致现存的推荐算法很难使用户关注长尾商品,影响了长尾商品的销售,如何提高对长尾商品的推荐显得十分重要.本文提出L_RRBM(Latent Dirichlet Allocation-Real Restricted Boltzmann Machines)算法,通过提取用户偏好及商品的主题,结合改进受限玻尔兹曼机对商品未知主题权重的预测,以解决对长尾商品的推荐问题.试验结果表明了本文推荐算法的有效性和可行性.

E-commerce platform has become an important way of selling goods,with the rapid development of it.However,the long tail effect of online merchandise sales has become an urgent problem to be solved in the research of electronic commerce.The evaluation of the popular commodity quantity is less,resulting in the existing recommendation algorithm is difficult to users concerned about the long tail of the long tail effect of goods,merchandise sales.At the same time,how to improve the recommendation of the long tail is very important.This paper proposes a L_RRBM（Latent Dirichlet Allocation-Real Restricted Boltzmann Machines）algorithm,through the extraction of user preference and commodity subject,improved prediction restricted Boltzmann machine for goods topic with unknown weights,in order to solve the problem of long tail products recommended.The experimental results show that the proposed algorithm is effective and feasible.