Test And Evaluation Approach For Network Application Systems Based On Data Fusion

With the coming of big data and network age, there are more and more networkapplication systems such as Internet sites, mobile APP and PC Software based on network, etc.The test of network application systems always focuses on representation and businesslogic.With the increasing scale of the network application systems, the architecture of them aremore and more complex. Single test tool can not test all the functions of the systems, butmultiple test tools are difficult to be integrated.Nowadays, the techniques of test for network application systems are not yet mature, testof network application systems is a open and key problem in the research and developing areasof computer network, and it has important theoretical significance and application value.In this thesis, the technique of broken links testing, the technique of connection speedtesting, the techniques of client and server testing are studied, and these techniques areverified and the test results are provided.Furthermore, in this thesis, the thought of fusingmultiple test results to further enhance the performance of the testing are proposed, and twofusion method is presented for network application systems.The first method is fusing test results of network application systems based on Bayesianclassifier network, continuous variables oriented Navie Bayesian classifier was adopted forfusion of test results, for example, if we fuse two test results which are represented as triples,aBayesian classifier including six attributes variables and one class variableclass can be used:the first three attribute nodes denote a test result, thereafter three attributes variables representthe second test results that was derived from some test method, class variable represents thefinal fusion result. The attribute nodes are continuous variables. Futhermore, the use ofensemble of multiple classifiers was discussed, that is to say ensembling multiple classifiers toform a hierarchical fusion framework in order to further improve the quality of fusion of testresults of the network application systems. In detail, we fuse multiple Bayesian classifier basedon affine propagation clustering.The second approach for fusion of test results of the network application systems is on thebasis of existing fusion method for ordered proposition, propose to use a simplified andmodified convex evidence theory to fuse the test results of network application systems.Experimental results demonstrate the effectiveness of the above two methods.