With　the　coming　of　the　time　of　Internet,　the　number　of　the　websites　is　growing　in　geometric　progression.　How　can　withhold　the　customers　and　how　to　make　websites　stand　out　above　the　rest　are　the　problems　in　managing　the　websites.　This　article　aims　to　analyze　the　behavior　of　web　users.　Firstly　log　data　were　collected　from　the　web　server,　after　data　preprocesssing,　a　URL-UserID　relevant　matrix　is　set　up　with　URL　as　row　and　UserID　as　column,　and　each　element　value　as　the　users＇　hits.　With　the　help　of　fuzzy　clustering　algorithms　and　Warshall　algorithms,　the　similar　customer　groups　can　be　discovered　by　measuring　similarity　between　column　vectors.　By　using　the　result　of　the　model,　we　could　analyze　the　behavior　of　users　and　improve　the　customer　relationship　management.　Finally,　Examples　of　the　web　log　data　of　Guangdong　Industry　Technical　College　prove　the　validity　and　effectiveness　of　the　algorithms.　©　2011　IEEE.