Adaptive Software Testing Based on Controlled Markov Chain

Adaptive testing for software means that software testing strategy should be adjusted online by using the testing data collected. Existing researches on adaptive testing rely on a simplified controlled Markov chain (CMC) model for software testing, which focus on how to detect and remove all of the defects with minimum expectation cost. The CMC model for software testing employs several unrealistic assumptions and this makes it have limited applicability. In order to overcome the limitations of applicability and low efficiency, this paper presents an improved CMC model with cost constraints by a series of new transformation of limit conditions. This model is designed according to the software cybermetics methodology. It can reach a balance among efficiency, complexity and applicability. Based on this model, a new optimal test strategy of software defects is designed to remove most defects within cost constraints. Then an adaptive testing strategy is proposed which can be adjusted on-line in accordance with the changes of parameters taking place in the controlled object. Finally a set of simulations are conducted and testing results prove the effectiveness of this model.