Abstract: Aiming at the status that users are difficult to obtain interesting information when accessing web resources, by analyzing users' requires, an optimized matrix clustering algorithm is proposed to improve web service quality. The user group and page are put in a cluster to understand the user behavior and find the users access patterns. Experimental results show that the method is feasible for more accurate reflection of the site access. According to the discovery of the browsing mode of users' access to the web page, the administrator can further analyze and study the regular pattern of web log records to improve Web site performance and organizational structure, so as to provide personalized service.