Bayesian Approach for Regression Testing (BART) using Test Suite Prioritization

Majority of the current state-of-the-art test suite (TS) prioritization algorithms for black-box testing focus on improving average percentage of fault detected (APFD) metric. These however suffer from two critical challenges 1) high time complexity of $\ge O(n^{2})$ where n is the number of test suites, and 2) limited ability to self-stop TS Prioritization (TSP) computation if the system under test (SUT) becomes highly stable. In this work we present an approach to overcome these two challenges while achieving high APFD efficiency over the conventional random ordering. A novel algorithm called Bayesian approach to regression testing (BART) is developed herein which models continuous integration (CI) cycle’s attributes like test suite life cycle (TSLC), stability and bugs as Bayesian inference pattern namely Dirichlet-Multinomial model. This work demonstrates that BART’s APFD metrics improve significantly in comparison to conventional random ordering and therefore this approach achieves for the first time a complexity of $O(nlogn)$ for black-box based test prioritization.