Partachi, Profir-Petru;
(2020)
Improving Software Project Health Using Machine Learning.
Doctoral thesis (Ph.D), UCL (University College London).
Preview |
Text
Improving_Software_Project_Health_using_Machine_Learning.pdf - Accepted Version Download (2MB) | Preview |
Abstract
In recent years, systems that would previously live on different platforms have been integrated under a single umbrella. The increased use of GitHub, which offers pull-requests, issue trackingand version history, and its integration with other solutions such as Gerrit, or Travis, as well as theresponse from competitors, created development environments that favour agile methodologiesby increasingly automating non-coding tasks: automated build systems, automated issue triagingetc. In essence, source-code hosting platforms shifted to continuous integration/continuousdelivery (CI/CD) as a service. This facilitated a shift in development paradigms, adherents ofagile methodology can now adopt a CI/CD infrastructure more easily. This has also created large,publicly accessible sources of source-code together with related project artefacts: GHTorrent andsimilar datasets now offer programmatic access to the whole of GitHub. Project health encompasses traceability, documentation, adherence to coding conventions,tasks that reduce maintenance costs and increase accountability, but may not directly impactfeatures. Overfocus on health can slow velocity (new feature delivery) so the Agile Manifestosuggests developers should travel light — forgo tasks focused on a project health in favourof higher feature velocity. Obviously, injudiciously following this suggestion can undermine aproject’s chances for success. Simultaneously, this shift to CI/CD has allowed the proliferation of Natural Language orNatural Language and Formal Language textual artefacts that are programmatically accessible:GitHub and their competitors allow API access to their infrastructure to enable the creation ofCI/CD bots. This suggests that approaches from Natural Language Processing and MachineLearning are now feasible and indeed desirable. This thesis aims to (semi-)automate tasks forthis new paradigm and its attendant infrastructure by bringing to the foreground the relevant NLPand ML techniques. Under this umbrella, I focus on three synergistic tasks from this domain: (1) improving theissue-pull-request traceability, which can aid existing systems to automatically curate the issuebacklog as pull-requests are merged; (2) untangling commits in a version history, which canaid the beforementioned traceability task as well as improve the usability of determining a faultintroducing commit, or cherry-picking via tools such as git bisect; (3) mixed-text parsing, to allowbetter API mining and open new avenues for project-specific code-recommendation tools.
| Type: | Thesis (Doctoral) |
|---|---|
| Qualification: | Ph.D |
| Title: | Improving Software Project Health Using Machine Learning |
| Event: | UCL (University College London) |
| Open access status: | An open access version is available from UCL Discovery |
| Language: | English |
| Additional information: | Copyright © The Author 2020. Original content in this thesis is licensed under the terms of the Creative Commons Attribution 4.0 International (CC BY 4.0) Licence (https://creativecommons.org/licenses/by/4.0/). Any third-party copyright material present remains the property of its respective owner(s) and is licensed under its existing terms. Access may initially be restricted at the author’s request. |
| Keywords: | project health, software engineering, applied machine learning |
| UCL classification: | UCL UCL > Provost and Vice Provost Offices > UCL BEAMS UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science > Dept of Computer Science |
| URI: | https://discovery.ucl.ac.uk/id/eprint/10116742 |
Archive Staff Only
 |
View Item |
