Explainable AI-Based Early Sprint Risk Prediction in Agile Software Projects Using Machine Learning Models

Abstract

Agile software development is widely used in modern software projects because it supports flexibility, continuous feedback, and iterative delivery. However, Agile projects still face several sprint-level risks such as reduced sprint velocity, frequent backlog changes, increasing defect count, poor task completion ratio, and excessive team workload. These risks can negatively affect sprint success if they are not identified at an early stage. Traditional risk management methods mainly depend on manual monitoring and project manager experience, which may not be sufficient for timely and accurate risk detection. Moreover, many existing machine learning-based prediction systems work as black-box models and do not clearly explain the reasons behind their predictions.

To address this problem, this research proposes an Explainable Artificial Intelligence-based early sprint risk prediction framework for Agile software projects. The proposed approach combines machine learning models with explainability techniques to predict sprint risk levels and provide understandable explanations for project managers. The system uses sprint-related features such as sprint velocity, backlog churn rate, defect count, task completion ratio, team workload, and story points completed. The target variable is sprint risk level, classified as Low, Medium, or High.

In this study, multiple machine learning models including Logistic Regression, Random Forest, Support Vector Machine, and XGBoost are applied and compared. The dataset is preprocessed through missing value handling, normalization, and encoding. Model performance is evaluated using accuracy, precision, recall, F1-score, and ROC-AUC. For explainability, SHAP is used to identify the contribution of each feature in the prediction process.

The key contribution of this research is the development of a transparent and interpretable sprint risk prediction framework that not only predicts early sprint risks but also explains the main factors responsible for those risks. This can help Agile project managers make timely, data-driven, and understandable decisions to improve sprint planning, reduce project failure, and enhance overall software project performance.