O2O Business Personalized Recommendation Research Based on Bipartite Graph

Bipartite graph networks are widely used as an analytical method for complex networks. Aiming at the cold start problem of the traditional recommendation algorithm, this paper proposes to combine the bipartite graph recommendation algorithm with the random walk algorithm, and then add the geographical influence factor to improve the effect of O2O ecommerce personalized recommendation. The experiment was conducted by selecting 45,974 users in the Yelp data set as an example. The online review was used as a weight to construct a user-business bipartite graph network and found to have a "small world" feature. The experimental results show that the combination of the random walk algorithm can make recommendations accurately, and add the characteristics of localization, by comparing the cities of the businesses, can be more in line with the actual situation. The accuracy of the final model algorithm is improved, which verifies the effectiveness of the model.