Investigation Of Data Mining-Based Software Testing

Software reliability is with great importance to social, economic, military and so on. To improve software's reliability, lots of tests must be implemented. But due to the restriction of time and cash, a balance point between cost and efficient must be found. How to achieve the best result with lowest costs has become an issue worth researching on. As a solution, Data mining has been introduced into software testing, and many papers have been published and methods invented in recent years.There exist two typical"Massive-Data"problems in Software Testing. One is How to choose test cases: As the input space of software is enormous, we can't test software using all these inputs. An efficient method is dividing the input space into several equivalence classes, and actual test cases are generated from these classes, which can significantly reduce the number of test cases. Another problem is analyzing Bug reports, many modern software are able to report Bugs automatically. It is an effective way to collect details of Bugs. But it always collects too much Bug report beyond developer's ability of analyzing. So we must find an automatic way to analysis these repots.In this paper, we give Data Mining-Based solutions to the problems above, they are: Using Info-Fuzzy Network to divide input space to several equivalence classes, and using Sequential Pattern Mining to analysis Bug Reports. They are proved to be available approaches by experiments...