Empirical Analysis of Greedy, GE and GRE Heuristics

Abstract

Whenever a software evolves, regression testing needs to be performed which ensures evolution does not affect the existing software. Test suite minimization is one of the regression testing techniques which takes a test suite and provides a minimized test suite (representative set) which is sufficient to cover all the requirements. This significantly helps in reducing the testing cost, effort and time. In this paper, we perform an empirical study on standard minimization techniques: Greedy, Greedy Essential (GE) and Greedy Redundant Essential (GRE), to compare their minimized test suite size (output size), and their run time to calculate the minimized test suite, with varying percentage of essential test cases. We assume that the number of requirements that a test case satisfies is a random variable which follows a normal distribution. This study analyzes trends in both output size and run time of the heuristics and enables us to set guidelines for the testing team to choose appropriate heuristics for performing the test suite minimization for all possible values of ratio of overlapping (defined as the average number of test cases which satisfy a requirement).