Explainable AI for Data-Driven Predictive Modelling Agile Projects

Abstract

Agile software development projects generate massive amounts of data related to tasks, sprints, team performance, defects, and delivery times. While machine learning-based predictive models are increasingly being employed to predict project outcomes like effort estimation, sprint velocity, risk, and delivery delays, the lack of transparency in these models often results in a lack of trust and acceptance among the project stakeholders. This paper explores the application of Explainable Artificial Intelligence (XAI) approaches to enhance the explainability of data-driven predictive models in agile project management. The proposed framework combines predictive machine learning techniques with explainability tools to generate clear and understandable interpretations of the model’s outputs. The model predicts the critical project matrix by analysing the historical data of agile projects from issue tracking systems and version control systems, and also explains how features such as team size, task complexity, sprint duration, and changes to the backlog impact the predictions. Techniques such as feature importance analysis and local explanation techniques in XAI are employed to ensure that the predictions are more interpretable and actionable.