Regression Test Predictor Using Unsupervised Learning Algorithm
SAE-PP-00174
08/26/2021
- Content
- Regression testing demands a precise and accurate test selection which, if done manually, requires thorough knowledge in functionality as well as expertise in testing. The execution time for regression testing is another challenge as the milestones for test activities are generally rigid. But lack of an effective regression testing leads to defect leakage and poor quality of the product. This study addresses the two major challenges of regression testing- expertise in test selection and time required for regression testing as a whole. A Regression Test Predictor Workflow Automation Framework is proposed which includes a machine learning based model for test selection. Appropriate tests are selected from the complete test suite by checking the syntactic and semantic similarity between the details of software changes and the testcases. As the input data changes every time, an unsupervised learning methodology is adopted for test prediction. The implementation of ML model is done for an automotive brake system software by * and the accuracy is calculated.
- Citation
- Mohanan, A., GS, S., and Menon, S., "Regression Test Predictor Using Unsupervised Learning Algorithm," SAE MobilityRxiv™ Preprint, submitted August 26, 2021, https://doi.org/10.47953/SAE-PP-00174.