Building a better recommender system in E-commerce

This paper presents an architecture of a hybrid recommender system in E-commerce environment. The goal of the system is to make special improvements in giving precisely personalized recommendation through some effective measures. Based on the study on the existing recommendation methods of both the conventional similarity function and the conventional feedback function, several improvement algorithms are developed to enhance the precision of recommendation, which include three improved similarity functions, four improved feedback functions, and adoption of both explicit and implicit preferences in individual user profile. Among them, issues and countermeasures of a new user, prominent preferences and long-term preferences are nicely addressed to gain better recommendation. The users preferences is so designed to be precisely captured by a user-side agent, and can make self-adjustment with explicit or implicit feedback.