In this paper, the linguistic evaluations and time duration on web pages are characterized as fuzzy variables in the web mining of preferred traversal patterns. The concept of fuzzy web page preference function which is a function of fuzzy linguistic evaluations, fuzzy time duration and access frequency is proposed to replace the traditional concept of confidence. The fuzzy web page preference function can be used as a judgment criterion to reveal users interest and preference. The structure of Frequent Link and Access Tree (FLaAT) which stores all user access information is introduced to avoid the loss of important information. Furthermore, an algorithm based on the fuzzy web page preference function is developed for mining users preferred traversal patterns from the structure of FLaAT. Finally an example is provided to illustrate the proposed approach.