Integrating AI and knowledge systems to automate bug reporting in software testing

Software (SW) testing is a vital part of software engineering, gaining importance as the field advances through key developments. Bug reporting analysis is a critical phase of SW testing. It requires collaboration between developers and clients. Manual bug review by experts often results in misclassification and delays due to the process’s complexity. Introducing the knowledge management (KM) cycle can enhance bug reporting by improving cost and quality. As the primary focus of this study is on classification and analytical processing of bug description to extract three key outputs (Bug Notification, Severity, and Responsible Side) that form the basis for bug reporting, this study presents the Testing Bug Report Generation (TBRG-KM) model; the basic idea for this is applying the KM principles to generate a learning-based system for bug reporting, analyzing and using Deep Learning (DL) as a technical implementation and embodiment for the TBRG-KM model. The TBRG-KM focuses on three key bug report features: Bug Notification, Severity, and Responsible Side. The model explores twelve variations using CNN, LSTM, GRU, and Bi-LSTM with attention mechanisms. The model was tested on a dataset comprising 20,171 bug reports from an agile e-commerce project targeting both web and Android frameworks. The Bi-LSTM with attention delivered the highest performance, achieving 0.95 accuracy for Bug Notification and 0.92 for Severity and Responsible Side.