Balancing accuracy and interpretability: AI-driven predictive modeling of construction schedule performance in India

Construction schedule delays are a persistent challenge in India’s infrastructure sector, leading to significant cost overruns, stakeholder dissatisfaction, and inefficient resource utilization. In this context, the integration of Artificial Intelligence (AI) into project forecasting offers a promising solution. This study presents an AI-driven predictive modeling framework aimed at estimating construction schedule adherence, with a dual focus on predictive accuracy and model interpretability. Three machine learning models—multiple linear regression (MLR), artificial neural networks (ANN), and gradient boosting machine (GBM)—were developed and evaluated using real-world data from 150 construction projects across India. The models were assessed based on standard performance metrics (MAE, RMSE, R²), computational efficiency, and transparency. While MLR offered high interpretability, it lacked accuracy in handling nonlinear interactions. ANN showed high accuracy but limited explainability. GBM outperformed both, achieving the best predictive performance (R² = 0.94) and a favorable trade-off between accuracy and interpretability. Feature importance and sensitivity analyses identified equipment utilization, material availability, and labor productivity as key influencers. Case study validation further reinforced the practical value of GBM in diverse project settings. The findings advocate for the strategic adoption of interpretable AI tools like GBM in construction management to enhance scheduling precision and decision-making efficiency.