AI-Powered Software Bug Tracking Tool

Abstract

Most information technology (IT) companies still rely on manual approaches to manage and track software bugs during testing sessions. This traditional method is timeconsuming, error-prone, and lacks efficiency particularly in large or complex software projects. To address these challenges, this paper proposes an artificial intelligence AI-powered Software Bug Tracking Tool designed to streamline the process of reporting, tracking, categorizing, and prioritizing bugs. By leveraging Artificial Intelligence (AI) technologies, including machine learning, the system can automatically classify bugs based on their severity, providing valuable insights for better decision-making. Additionally, the system employs unsupervised learning techniques to detect patterns and relationships among bugs. An OpenAI GPT-4o model, accessed via the Azure OpenAI Service, is integrated into the system to automatically generate summaries of bug reports, helping team quickly understand the key details. A user-friendly design is also provided to support collaboration between developers and testers, reducing the learning curve and improving overall software quality. For bug severity detection, the F1-score is 0.64 when detection Critical bugs; an indicator that the approach might be useful. Similarity bug detection is more challenging because the similarity scores might not necessarily reflect the bug similarity. Test case summarization is based on the OpenAI API. The implementation of this system aims to enhance team productivity, ensure comprehensive bug tracking, and align with modern software development practices. Based on the usability testing with 10 participants, including software expert from various background such as software testers and developers, the duplicate detection and bug summarization features are particularly useful, to streamline the bug tracking process and improve the understanding of reported issues. However, the bug severity prediction feature produced unexpected classifications in certain cases, indicating the need for additional training data and the potential future works to enable the model to continuously improve over time.