Design of Product Recommendation System based on Restricted Boltzmann Machine

The traditional recommendation system is to use the evaluation of products by neighbors with high similarity to the target user to predict how much the target user likes the product, but its drawback is that the degree of individual user profiling is not enough, and the recommendation accuracy should be improved again. In this paper, we integrate the RBM model with clustering, which can optimize the shortcomings of traditional recommendation models. The restricted Boltzmann machine is a graphical model of binary random variables, based on a complete bipartite graph separating hidden and observed variables, and is a binary simulation of the factor analysis model. The RBM model can fully consider the connection between users and products, and the clustering algorithm focuses on the connection between users, and combining them both can improve the accuracy of the recommendation system