A Mathematical Modelling Approach for Software Testing Optimization

Software errors can be a serious problem, because of possible dam- ages (and related costs) and the burden of the needed corrections. Software testing, whose aim is to discover the errors in software products, requires a lot of resources and from it derives the overall quality (i.e. reliability) of the software product. It is thus susceptible of optimization: i.e., the best equilibrium between the number of tests to make and the global expected value of discovered errors has actually to be achieved. In fact, it is not economically feasible to proceed with testing over a given limit as well as to execute too few tests, running in the risk of having too heavy expenses because of too residual errors. In the present work we propose some models for software testing optimization, making use of an integer linear programming approach solved with a “branch & bound" algorithm