Test process optimisation through big data analysis
The goal of the research is to adopt artificial intelligence techniques for regression test case selection in an industrial setting. Currently, the selection of test cases is made manually by test engineers referring to software change documentation. A machine learning-based solution which selects t...
| Main Author: | |
|---|---|
| Format: | Thesis (University of Nottingham only) |
| Language: | English |
| Published: |
2023
|
| Subjects: | |
| Online Access: | https://eprints.nottingham.ac.uk/74011/ |
| _version_ | 1848800834900459520 |
|---|---|
| author | Kho, Xiang Juan |
| author_facet | Kho, Xiang Juan |
| author_sort | Kho, Xiang Juan |
| building | Nottingham Research Data Repository |
| collection | Online Access |
| description | The goal of the research is to adopt artificial intelligence techniques for regression test case selection in an industrial setting. Currently, the selection of test cases is made manually by test engineers referring to software change documentation. A machine learning-based solution which selects test cases with high potential in uncovering software defects is proposed and results were analysed.
Past historical results of test cases and various metadata, such as the ID, name and priority of test cases, were used in data training. At the pre-processing stage, data were analysed, cleaned and normalised. Then, a two-part balancing method, comprising of outliers removal and resampling with an algorithm, was applied to the imbalanced data before it can be fitted to various machine learning models.
The model that best fits the system requirement, which recommends at most 50% of the total test cases with no false negative predictions and as few false positive predictions as possible, is selected to be implemented as an executable application. The finalised model, based on random forest, recommends 1,626 test cases (7.35% of total test cases) for execution with no false negative and 882 false positive predictions (3.98% of total test cases) out of 22,137 test cases. This fulfils the two objectives of this research, which is to construct a recommendation system capable of reducing at least 50% of all test cases, and to ensure that there are no false negative and minimal false positive predictions.
The current test system is a black-box system, meaning the software’s functionalities are tested without accessing its internal code structure. The model’s performance can be improved with a white-box test system, where the software source code can be accessed, and information such as which specific code segments are causing test case failure is available. |
| first_indexed | 2025-11-14T20:57:52Z |
| format | Thesis (University of Nottingham only) |
| id | nottingham-74011 |
| institution | University of Nottingham Malaysia Campus |
| institution_category | Local University |
| language | English |
| last_indexed | 2025-11-14T20:57:52Z |
| publishDate | 2023 |
| recordtype | eprints |
| repository_type | Digital Repository |
| spelling | nottingham-740112025-02-28T12:27:26Z https://eprints.nottingham.ac.uk/74011/ Test process optimisation through big data analysis Kho, Xiang Juan The goal of the research is to adopt artificial intelligence techniques for regression test case selection in an industrial setting. Currently, the selection of test cases is made manually by test engineers referring to software change documentation. A machine learning-based solution which selects test cases with high potential in uncovering software defects is proposed and results were analysed. Past historical results of test cases and various metadata, such as the ID, name and priority of test cases, were used in data training. At the pre-processing stage, data were analysed, cleaned and normalised. Then, a two-part balancing method, comprising of outliers removal and resampling with an algorithm, was applied to the imbalanced data before it can be fitted to various machine learning models. The model that best fits the system requirement, which recommends at most 50% of the total test cases with no false negative predictions and as few false positive predictions as possible, is selected to be implemented as an executable application. The finalised model, based on random forest, recommends 1,626 test cases (7.35% of total test cases) for execution with no false negative and 882 false positive predictions (3.98% of total test cases) out of 22,137 test cases. This fulfils the two objectives of this research, which is to construct a recommendation system capable of reducing at least 50% of all test cases, and to ensure that there are no false negative and minimal false positive predictions. The current test system is a black-box system, meaning the software’s functionalities are tested without accessing its internal code structure. The model’s performance can be improved with a white-box test system, where the software source code can be accessed, and information such as which specific code segments are causing test case failure is available. 2023-07-22 Thesis (University of Nottingham only) NonPeerReviewed application/pdf en arr https://eprints.nottingham.ac.uk/74011/1/MPhil%20Thesis%20-%20Kho%20Xiang%20Juan%20%2820220157%29.pdf Kho, Xiang Juan (2023) Test process optimisation through big data analysis. MPhil thesis, University of Nottingham. artificial intelligent; big data analysis; software development; advance technology |
| spellingShingle | artificial intelligent; big data analysis; software development; advance technology Kho, Xiang Juan Test process optimisation through big data analysis |
| title | Test process optimisation through big data analysis |
| title_full | Test process optimisation through big data analysis |
| title_fullStr | Test process optimisation through big data analysis |
| title_full_unstemmed | Test process optimisation through big data analysis |
| title_short | Test process optimisation through big data analysis |
| title_sort | test process optimisation through big data analysis |
| topic | artificial intelligent; big data analysis; software development; advance technology |
| url | https://eprints.nottingham.ac.uk/74011/ |