Integrated Risk Analysis and Delay Prediction in Construction Projects using XGBoost, Monte Carlo Simulation, and Rule-Based Reasoning

Abstract

Infrastructure development relies heavily on timely construction project execution; however, schedule delays remain a critical challenge that escalates costs and compromises quality. This study develops an intelligent decision support system integrating qualitative risk assessment with quantitative predictive modeling to analyze and mitigate schedule delay risks. A comprehensive risk matrix survey of 25 construction professionals identifies weather conditions as the most dominant delay factor (risk score: 18.3), establishing the empirical foundation for the system architecture. The framework utilizes a unified dataset of 11,000 records combining daily project operational data—including progress, work volume, labor deployment, material receipts, and equipment usage—with meteorological data from BMKG covering 2021–2023. Data preprocessing encompassed cleaning, feature engineering, and SMOTE method to address class imbalance. The Extreme Gradient Boosting (XGBoost) algorithm classifies daily task feasibility as safe, at-risk, or delayed. Monte Carlo simulation with 10,000 iterations quantifies duration uncertainty due to weather variability and productivity fluctuations, while a Rule-Based Reasoning (RBR) module generates automated mitigation recommendations including rescheduling and resource reallocation. Evaluation using Repeated K-Fold Cross Validation demonstrates robust performance: 95.6% accuracy, 93.2% precision, 94.1% recall, and 93.7% F1-score. Monte Carlo analysis reveals schedule variations between −10% and +25% relative to planned durations. The RBR module successfully reduces average project delays by 21.5% and improves labor efficiency by 14.2%. These findings demonstrate that integrating risk analysis, machine learning, and stochastic simulation within a unified computational framework significantly enhances predictive accuracy and enables proactive delay mitigation, supporting adaptive decision-making under dynamic weather and operational conditions.