User modeling for personalized recommender systems

This paper models a user' interest and value related characteristics based on the extension by using neural networks and econometric techniques to the concept of user model in recommender systems. A two-dimensional taxonomy of user model was proposed in terms of the content and persistence of user characteristics. A user interest model and a customer lifetime value model were developed for the proposed taxonomy framework, to capture the time-dependent evolving nature of user's interests and his/her long-term profitability. The proposed models were empirically validated by using real customer data from a bee product company in health care industry. The experimental results show that these models provide effective assistant tools for the company to target its most valuable customers and implement one-to-one personalized services.