An Improvement to Test Case Prioritization Techniques Using Machine Learning

Abstract

AbstractRegression Testing is an important process every time new changes are integrated in the software product. Managing test cases becomes important in this scenario, where testers need to retest or select test cases based on the changes incorporated. Machine learning can reduce the effort of testers by providing a simplified approach in test case management. Prioritization of test cases can be taken as an initial step of handling bulk test cases and arranging them on the basis of certain priority. Several methods were proposed in the past that handle test cases in case of regression testing. This paper proposes a novel methodology that uses unsupervised machine learning algorithm to prioritize the test cases. Methodology contributes toward risk-based approach whose objective is to prioritize the test cases by identifying the most risky areas first. It categorizes the test cases in clusters representing different priority level. Our proposed approach is based on four essential factors namely the Change-Module mapping, Degree of Coupling between effected modules, Test Coverage Reports and Test Execution Results. Research objective relies on exploring how dynamic change information when combined with white box testing parameters and unsupervised machine learning can be beneficial in categorizing test cases. Paper also presents analysis of the existing literature related to test case prioritization, identified research gaps and expected contribution toward the literature.KeywordsRegression testingTest case prioritizationChange impactRisk-based testing