Researching Of Web Log Statistical Testing Based On User Classification

Since Web application can be characterized as massive userpopulation, diverse usage environments and document or informationfocus, the traditional software testing is facing new challenges. Becauseof the multilayer architecture of Web applications, different kinds ofcomponents can be used between layers, and coverage testing cannot do well in the case of limited recourses. In this paper, a method ofstatistical Web testing, strategy for testing focus on the key modules, isproposed to meet the demand. Statistical testing is aimed at testingfrequently usage of pages and paths, and we can use theperformance of the subset of software to assess the reliability for thewhole software.The usage model should be constructed according to the realusage of the Web application. UML based usage model and Markovchain based usage model are two important usage models forstatistical testing. As Web applications become more complex, thetraditional two models have been unable to adapt to the new Webapplications which functions and user permissions are staggered. In thispaper, according to a hierarchical Markov chain model, we propose anew approach of statistical testing based on user classification. ThroughWeb log analysis and extraction, the user of Web application can beclassified by this approach and the Markov chain model based on theclassification can be constructed for testing.“Campus IT services platform” is used as a test platform in thispaper. We illustrate that how to separate users from the Web access log,and how to construct the usage model using different algorithms; alsowe apply this method into the test platform and the results demonstratethis approach is feasible. This approach can adapt to higher complexof Web applications and makes the assessment of reliability for Web applications more viable and realistic.