The concept of preference is proposed on the analysis of the present algorithms for mining user navigation patterns. It is based on the following hypothesis: if there are many different selections to leave a page, those selections that occur more frequently and the next page is viewed longer reveal user interest and preference. Representing user navigation interest and intention accurately by comparing relatively access ratio and the average of relatively access ratio of viewing time and selective intention, preference can be used for mining user navigation pattern instead of confidence. The higher preference, the more prefer to choose the selection. According to the conception, we propose two efficient algorithms based on the concept, UAM and PNT, which are developed for mining user preferred navigation patterns. Considering the structure of Web site, UAM can get user access preferred path by the page–page transition probabilities statistics of all users behaviours. PNT looks far into the past to correctly discriminate the different behavioral modes of the different users. Experiments show accuracy and scalability of the algorithms. It is suitable for applications in E-business, such as to optimize Web site or to design personalized service.