AI-Based Schedule Overrun Prediction in Technical Projects Using Gradient Boosting Machine for Decision Support

Abstract

The increasing complexity of technical projects has challenged the effectiveness of traditional project management approaches, particularly in accurately predicting schedule performance. Conventional methods often rely on expert judgment and static planning techniques, limiting their ability to capture dynamic project conditions and complex relationships between variables. This study investigates the role of Artificial Intelligence, specifically the Gradient Boosting Machine (GBM), in enhancing schedule management through a data-driven approach. A predictive model was developed using a publicly available project management dataset obtained from Kaggle, consisting of 4517 project instances. The dataset includes key variables such as project duration, total cost, team size, risk factor, and client satisfaction. The proposed model was applied to predict Schedule Overrun (%) as a measure of project performance. The results demonstrate strong predictive performance, with an R² value of 0.835, RMSE of 7.128, and MAE of 5.946. These findings indicate that the model is capable of capturing complex patterns and relationships within project data. Feature importance analysis revealed that risk factor and project duration are the most influential variables affecting schedule performance, followed by client satisfaction. These results highlight the importance of managing project uncertainty and timeline characteristics to improve overall project outcomes. The study confirms that AI-based models can provide valuable predictive insights to support proactive, data-driven decision-making in project management. Future research may focus on expanding the dataset, incorporating additional contextual variables, and exploring advanced modeling techniques to further enhance predictive accuracy and practical applicability.