Dual Preference Matrix Collaborative Filtering Algorithm and Its Application in Railway B2BE-commerce Platform

摘要

The enterprise users of the B2B e-commerce platform are different from the individual users of B2C, and their behavior has typical features of enterprise. Based on the original user-commodity matrix of collaborative filtering algorithm, the enterprise-category matrix which manifests the features of enterprise users is added to ameliorate and form a dual preference matrix concerning both user-commodity and enterprise-category relationship, thus can be applied to Railway B2BE-commerce platform. The experimental results indicate that the improved dual-preference matrix collaborative filtering algorithm has advantages over the original collaborative filtering algorithm and KNN algorithm in terms of accuracy, recall rate, coverage rate and degree of novelty, effectively solving the problem of cold start and improves the accuracy of recommending commodities towards inactive users by over 10%.