Research Of Software Testing Strategy Based On Markov Model With Defects Correlation

With the growing scale of the software development and the increasing complexity of thestructure, software reliability problems have become increasingly prominent. Software testingthat is the most commonly used method to ensure software system quality is increasingly playingan important role in software engineering. The Markov chain testing model has become animportant research direction of software testing. This model abstracts the applications into atesting model and analyze the testing model to select the appropriate testing decisions. Theunderlying motivation is to formalize, quantify, and optimize the software testing process. Alarge number of researchers conducted in-depth research and made a lot of outstandingachievements in software testing. However, most of these studies assumed that all defects areindependent of each other, i.e., ignoring the relationship between defects. In the practical testingprocess, yet defects are normally correlated each other. Therefore, this paper focuses the study onhow to use the information between correlated defects to improve Markov testing model andoptimize the software testing strategy.From the standpoint of the defects themselves, this paper studies in terms of the correlateddefects, the testing model based on Controlled Markov Chain and software testing strategy,including the following aspects:(1) The Markov testing model with defects correlation. Aim at the limitations of theexisting Markov model, this paper introduces an correlated defects coefficient matrix and refinerebate mechanism to construct a Markov model based on defects correlation. So that the modelcan synthetically balance the weights design based on defect correlation coefficient, defectdetection rate and rebates to improve the testing efficiency.(2) The software testing strategy based on Markov model with defect correlation. Bymaking full usage of relationships between correlated defects, this paper converts the softwaretesting process into a weighted routing problem and designs a software testing strategy based on a Minimum Spanning TreeAlgorithm to obtain an optimized testing path with maximum weights.Meanwhile, we research a more complex correlated defect cases and change the testing modelinto a directed weighted graph problem. Moreover, An improved Prim Algorithm is proposed tosupport this model.(3) Batch removal strategy. A new batch removal strategy is proposed to effectivelyalleviate the failure shielding problem, which defects detected capability masked by other defectsdue to prematurely eliminate a single correlated defect.In order to verify the validity of the test model and test strategies, a set of simulations areperformed. Experiment results show that our approach has higher effectiveness in terms of thenumber of detected defects, testing cost and rebates.