Assessing two graph-based algorithms in a model-based testing platform for Java applications

Abstract

Model-based testing (MBT) is an approach for automatically generating test cases from a model of the system under test. Existing MBT tools support the automation of this process at varying degrees. One such tool is MBT4J, a research platform that extends ModelJUnit, offering a high level of automation. We extended MBT4J with two graph-based algorithms: the Chinese Postman Problem (CPP) and Breadth-First Search (BFS). The purpose of this study is to evaluate the efficacy of these two new algorithms added to MBT4J by comparing them to previous algorithms implemented in the platform. A case study was conducted using two open-source Java applications from public repositories, and twenty-one different configurations. The CPP tester performed similarly to previous testers in terms of time and coverage, and in addition, it resulted in a greater percentage of failed test cases in one application. The BFS tester was able to generate a greater amount of test cases when using fewer resources. We thus recommend using these algorithms for generating test cases for systems with complex models.