Bug Triage Model Using Machine Learning Techniques
| dc.contributor.advisor | Belay, Ayalew (PhD) | |
| dc.contributor.author | Andualem, Meseret | |
| dc.date.accessioned | 2021-11-16T07:08:12Z | |
| dc.date.accessioned | 2023-11-29T04:06:35Z | |
| dc.date.available | 2021-11-16T07:08:12Z | |
| dc.date.available | 2023-11-29T04:06:35Z | |
| dc.date.issued | 2021-08-27 | |
| dc.description.abstract | A customer, analyst, or developer makes an error while using the system or generating software artifacts, which is a common chain of reaction to software bugs. This error may cause a system fault, resulting in an unexpected state. This unanticipated situation could lead to a bug, which is a visible and unwelcome event from the user's perspective. When a bug is detected, the user creates a bug report that includes the error messages issued by the software. These bugs must be dealt with in a proper manner. Bug triage is one of them. Triaging is the process of categorizing and prioritizing bug reports in order to assign priority and route them to the appropriate developer for resolution. Many researchers have developed a number of bug triage models in recent years. However, because bugs in bug repositories aren't necessarily bugs, new strategies to identify the actuality of the bug are still needed. As a result, we present a bug triage model that uses machine learning techniques to identify actual bugs from non-bugs by assessing the severity field and then assigning to the right developer based on the developer's tossing history. Preprocessor, feature extractor, dataset constructor, bug detector, and bug assigner are the components of the suggested model. Data tokenizing, stop word removal, and stemming are the main operations in the preprocessing component. The feature extraction component then extracts the feature vectors from the supplied bug report, while the dataset constructor splits the dataset into training (80%) and testing (20%) sets by converting the javascript object notation (JSON) file into sets. Finally, based on the recognized component retrieved from the bug report's short description, the MNB classifier is used to classify the bug report into BUG and NON BUG and automatically propose developers who have the necessary competence for processing a bug report. The feasibility of our proposed model has been validated on Eclipse covering 315228 bug reports. We show that our techniques can detect bugs and assign them to them with a prediction accuracy of up to 95.67 percent. | en_US |
| dc.identifier.uri | http://etd.aau.edu.et/handle/123456789/28688 | |
| dc.language.iso | en | en_US |
| dc.publisher | Addis Ababa University | en_US |
| dc.subject | Bug Triage | en_US |
| dc.subject | Bug Detection | en_US |
| dc.subject | Bug Assignment | en_US |
| dc.subject | Multinomial Naive Bayes | en_US |
| dc.title | Bug Triage Model Using Machine Learning Techniques | en_US |
| dc.type | Thesis | en_US |