E-Commerce facilitates our daily life and is playing a very important role in today's business. With the development of software and network technique, people have more and more request on E-Commerce especially for the non-mainstream business. Individualized identification is one of the most important supporting for the clients who want to purchase some non-mainstream goods from web. In this paper, we capture the geographic information embed in the historic web operations for some registered individuals and suggest using hidden Markov model to identify their preference when they purchase the non-mainstream product (or services). We have established a prototype and the results turned out to be wonderful in the pilot.