With the rapidly development of Internet, online medical platform has become an essential part of medicines trade. In order to help users quickly find satisfying products in a large number of commodities, the recommendation system has been proposed. The traditional recommendation algorithm usually only takes the user-item rating into consideration, which leads low accurate of prediction. In this paper, we propose a user profile based recommendation method, which uses deep learning to analyze user behavior and construct user multi-dimensional attribute features. user profile can be constructed by analyzing information of drugs. By analyzing the historical information of user's action, including purchasing, browsing, and collecting, we can dynamically predict rating of user on drug by a trained neural network. The experimental verification on B2B medical platform shows that the accuracy of prediction is higher than other algorithms. The proposed system can not only improve user experience, but also increase the sales of the platform.
展开▼
