A Fuzzy Expert System for Cost-Effective Regression Testing Strategies

Different testing environments and software change characteristics can affect the choice of regression testing techniques. In our prior work, we developed adaptive regression testing (ART) strategies to investigate this problem. While the ART strategies showed promising results, we also found that the multiple criteria decision making processes required for the ART strategies are time-consuming, often inaccurate and inconsistent, and limited in their scalability. To address these issues, in this research, we develop and empirically study a fuzzy expert system (FESART) to aid decision makers in choosing the most cost-effective technique for a particular software version. The results of our study show that FESART is consistently more cost-effective than the previously proposed ART strategies. One of the biggest contributors to FESART being more cost-effective is the reduced time required to apply the strategy. This contribution has significant impact because a strategy that is less time-consuming will be easier for researchers and practitioners to adopt, and will provide even greater cost-savings for regression testing sessions.