Modeling customer preference for E-commerce recommendation

Customer preference is a relation of a customer and a product. Usually it is represented by the set of attributes in order to predict the preference of new products, and the actual value is estimated from the customer history record. Therefore, customer preference model is required in intelligent E-Commerce recommendation systems. In this paper, we apply joint product attribute and dynamic weighting to model the customer preference and attribute preference. Pareto distribution and random probability are employed to reduce effects caused by data sparseness problem. The experimental results show that our preference models can effectively improve the recommendation precision.