Predictive regression test selection technique by means of formal concept analysis

Regression testing is an important software maintenance process that is applied to validate the correctness of software modifications. Since regression testing is usually costly, selective regression testing can be an alternative to traditional regression testing, allowing for a reduction of the overall cost associated with testing. Selective regression testing identifies only the test cases that execute parts of a program that can potentially be affected by the modification. In this thesis, we propose a novel technique to perform selective regression testing by means of a data analysis technique, Formal Concept Analysis. We use the capability of Formal Concept Analysis to structure the commonalities of execution traces derived from existing test cases. Formal Concept Analysis provides information related to program execution dependency among different parts of a program, which can then be used to determine the relationships between a modified program portion and existing test cases. In this research, a novel approach analyzes the program execution dependency to allow for the selection of test cases that should be rerun after the program modification is complete. We have implemented a tool that automates regression test case selection and demonstrates a proof of our concept.