Recommendation and Regression Test Suite Optimization Using Heuristic Algorithms

Abstract

In the Software Development Life Cycle, testing is an integral and important phase. It is estimated that close to 45% of project cost is marked for testing. Defect removal efficiency is directly proportional to the rigor of the testing and number of test cycles. Given this prelude, important optimization dual is to reduce the testing time and cost without compromising on the quality and coverage. We revisit this popular research and industry sought problem, in the historical data perspective. For this, it is important to follow an approach and minimize the available test suites and recommend N Test cases based on multiple heuristics. The heuristics can be derived based on Test Manager, Test Lead and/or Test Director requirements and inputs. The N test cases that are to be recommended will be derived upon executing evolutionary randomized algorithms such as Random Forest / Genetic Algorithm. These algorithms fed with historically derived inputs such as test case execution frequency, test case failure pattern, change feature pattern and bug fixes & associations. The recommended test suite is further optimized based on a 2 dimensional approach. Test case specific vertical constraints such as distribution of environments, distribution of features as well as Test suite composition parameters such as golden test cases, sanity test cases, that serves as horizontal parameters.