Abstract: This paper presents the coefficient of users' sessions AvgI (p) according to the problem for neglecting the users' level of interest in the page in some session, which is in the calculation of users' session similarity.Also, it introduces a new calculation method for the level interest of users' sessions, combining Jacobian similarity coefficient, CM coefficient.On this basis, combining with Web content and format, the authors give its algorithm through further analysis of the information needs of the user intention and classification of two types of navigation, i.e., direct navigation and indirect navigation.Experimental results show that the algorithm can automatically acquire users' search intentions under navigation, and provide relevant results and further navigation based on the intentions.