Web log mining is the most important application in Web data mining. We can improve the organization structure of Web site and its function ,increase personalized service and discover the potential reader group on the basis of the analysis and research of  Web log mining documents. Data preprocessing decides the quality of Web log mining. It includes data clearing, user identifying, user session identifying, format, etc. and its aim is to separate Web server log into multi-user reference strings and also give the reference type realization.