Research And Implementation On E-commerce Recommendation System

With the adventure of E-commerce applications, personalization has assumed an enormous industrial impact, which has caused the appearance of a menagerie of commercial products, claiming support to the personalization process. Personalization demonstrates the potential of intelligent information services that are capable of automatically learning about the short and long term needs of users, and of customizing the search and delivery of information to these users. Current recommendation methods expose some quality problems, such as Data Sparsity problem and New-Item problem, these problems make ordinary recommendation systems can not mine users'interests and provide most proper items to them.In this paper, we describe a new recommendation method, in this method, a product Semantic Relevance Model is set up, the structure of this model is acquired by the training of an Industry Semantic Training Center, finally semantic information that extracted from products is integrated with current Item-Based Collaborative Filtering approach. The method makes use of characteristic of Item-Based Collaborative Filtering approach, which improves the extension and precision of recommendation information through behaviors of other users, and combines Semantic Filtering approach, which can deeply mine users'connotative interests. In this paper, product information and user preference information are demonstrated in RDF forms, and a new user Semantic Preference Model which is based on decision tree is designed, the model is setup off-line, applied on-line, updated periodicly, which improves system real time performance effectively.Finally, we implement the recommendation algrithems, design and develop an E-commerce Recommendation System of Integrated Semantic Information, which is based on B/S pattern and provides a series of operations to users and administrators.