With the growth of the Internet, World Wide Web increasingly helps people to make good use of rich information from local or remote site. The amount of Web page s is so enormous that users are destined to drown in the huge data of the Web wi thout any navigation. To provide a new navigation approach, a modified Markov ch ain model is introduced, which utilizes all group members' traces in the Web to recommend some potential useful Web sites and navigates people when they browse Web pages, while users' activities react to the model. Before that, an algorithm based on a semi-formal description of process is necessarily given for collecti ng desired data to gain top grade results. The method is also illustrated by ana lyzing the proxy server's access log in the prototype system.