Heterogeneous information network to support the bug report resolution process

• Ano de defesa: 2021
• Autor(a) principal: Barbosa, Jacson Rodrigues
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Aprendizado de máquina; Bug report resolution process; Heterogeneous information network; Machine learning; Mineração de documentos de software; Processo de resolução de relatórios de incidentes; Rede de informações heterogêneas; Software documents mining
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/55/55134/tde-02022022-160727/

Context. Throughout a softwares lifecycle, numerous documents (e.g., bug reports and source code) are produced by stakeholders. Bug reports (BR) are the primary input documents to support the activities (bug report severity prediction and fixer recommendation) of the bug report resolution (BRR) process. Source code combined with bug reports are resources to support troubleshooting activities. Automation of these activities of the BRR process requires a concern with obtaining a semantically representative representation. Traditionally, Bagof-Word (BoW) represents software documents to support the automatic execution of these activities through machine learning algorithms. Gap. However, little attention has been paid to representations based on heterogeneous information networks (HEN), which allow representing complex networks respecting the relationships between different objects. Contribution. This doctoral thesis contributes to advancing state of the art regarding information representation models to support the automatic execution of activities in the BRR process. It also advances in the investigation of (i) semi-supervised algorithms that use bipartite heterogeneous networks to support the bug report severity prediction, (ii) a method that combines the BoW representation and heterogeneous information networks to support the bug localization activity, and (iii) a holistic approach that reuses a heterogeneous information network to support BRR activities. Results. The results demonstrate that heterogeneous information networks can be a promising alternative to support the automation of the BRR process. Conclusions. An automatic BRR process using a heterogeneous information network in a holistic perspective presented promising results compared with state-of-the-art representations.