| With the coming of the Internet age, the information is in the explosive growth. It ishard to quickly find what people need in a flood of information items. In this context,personalized recommendation technology comes into being. However, the most widelyused Collaborative Filtering recommendation technologies have some problems, such ascold start, data sparsity. Those affect the accuracy of recommendation results. As a result,the label technology, which is popular in web2.0, is introduced into the personalizedrecommendation system. Meanwhile, in order to improve the accuracy of therecommendation systems, we take user’s score, scoring time, and the trust factor intoaccount, creating an e-commerce personalized recommendation system based on trust andusers’preferences. Major studies are summarized as below: Ona measure of user interest,the existing recommendation technologies often only consider one or two factors. However,as important properties to reflect user interest, user tags, user ratings and rating time are allvery important. Therefore, it’s necessary to consider all those factors synthetically, andthus produces more accurate recommendation results. Based on this, we suggest an itemrecommendation model which is considering tags, the item rating and timeliness of user’spreference all together. This system can improve the accuracy of recommendation resultseffectively. Theoretical research shows that trust has positive relationship with userpreference’s similarity. Based on Multi-factor model, trust is brought into the new system,and we suggest an e-commerce personalized recommendation system based on trust andusers’ preferences. At the same time, we do the research on personalized itemrecommendation through direct and indirect sides separately. The model is morereasonable in producing neighbor users. Because compared to user interest informationcontaining timeliness of user’s preference, using trust to measure user ’s similarity is moreaccurate; meanwhile, this model contains the advantages of the model integrated tags, theitem rating and timeliness of user’s preference in measuring single user’s interest, andprovides a solution in the situation of cold start, black box and data sparsity which are themost important problems in Collaborative Filtering recommendation systems. |
