Regression Test Suite Optimization using Adaptive Neuro Fuzzy Inference System

Regression testing is an important activity performed to ensure that changes in the baseline version of the system do not influence the already tested part of the system. It becomes difficult to run the entire test suite due to constrained or limited resources. A subset of test cases that is as efficient as the original test suite is searched as optimal suite. Computational intelligence approaches has been used to search the representative subset. The major concern is safe reduction of test suite. Safe reduction has been achieved by fuzzy optimization. Neural networks are known for their ability to learn and fuzzy based systems for their quality to judge and make decisions. Neuro-fuzzy systems can be used to learn and make expert judgment. We have implemented Adaptive Neuro-Fuzzy Inference System (ANFIS) for test suite optimization. The resultant suite was found to be better than fuzzy based optimization in reducing the time and improving the coverage of resulting test suite. We concluded that ANFIS can be used to automate the optimization process. We will implement it on sufficiently large sized test suite available in software infrastructure repository provided by Siemens to validate our findings.