Applying customer-centered recommendation on an on-line shopping system

One of most popular applications developed for electronic commerce (EC) is the Recommender Systems (RS) which is usually built within the vender´s website. Since relevant customer information is needed to enable RS to provide personalized recommendations, it is necessary for customers to provide their personal information every time they visit a new e-store. Consequently, the more websites customers visit, the more likely their personal information is disclosed. Under such circumstance, not only the capability of the RS is reduced, but the risks of customer privacy violation are increased. This research proposes a new three-tier RS architecture including the customer tier, the third-party server tier and the vender e-store tier based on the customers´ perspective. In the proposed RS, customer information and recommendation module are built in the customer tier. By the help of the third-party server, cross-store recommendation is provided from the customer relative information. At the same time, the new structure also overcomes the problem of “poor new-user recommendation”, and reduces the risk of privacy disclosure. Base on the proposed architecture, a new prototype system is developed and evaluated through laboratory experiments and questionnaires. The results demonstrate that the participants were significantly more satisfied with the proposed RS architecture than the traditional ones.