Inteligência artificial & metodologias ágeis: como aprimorar o Confidence Vote em PI Plannings utilizando redes neurais

Abstract

This undergraduate thesis was developed as part of the Computer Engineering program at UTFPR, Curitiba campus, and explores the intersection of two highly important areas for modern software project development: Artificial Intelligence (AI) and Agile Methodologies. The central focus is on improving the Confidence Vote in PI Plannings, a common practice in Agile Methodologies for task prioritization and effort estimation. To achieve this goal, the study proposes the application of Artificial Neural Networks (ANNs), a subfield of AI, as a decision-support tool during the Confidence Vote. The ANNs were trained on historical project data, learning to identify patterns that contributed to better resource allocation and task prioritization in iteration planning (PI Plannings). The research involved analysis, experimentation, and evaluation of the proposed approach’s effectiveness. It is expected that the integration of AI with Agile Methodologies will lead to a significant improvement in the accuracy of effort estimates and, consequently, an increase in the quality and productivity of software development projects. This thesis aims to provide collaborative contributions to the software engineering community and Agile practitioners by exploring how AI can be innovatively applied to optimize planning and prioritization processes, making them more efficient and effective. Additionally, it reinforces the importance of continuous evolution and adaptation in software development practices within an ever-changing technological landscape.