BIM-Integrated Machine Learning Approach for Interactive Cost and Schedule Risk Prediction and Optimization in Urban Building Construction

Construction projects in urban areas, especially in developing countries, frequently experience significant cost overruns and delays due to uncertainties in resource allocation, labor productivity, and various risk factors. Traditional project management methods lack dynamic predictive capabilities and interactive decision support. Therefore, this study aims to create a building information modeling (BIM) based machine learning (ML) method for cost and schedule risk prediction and optimization for urban construction projects.

In order to provide a reliable quantity takeoff and a detailed baseline schedule, a real four-story educational building (1250 sq ft) was modeled in Revit. All quantities were transferred into Primavera P6 for cost estimation. A dataset of 1200 instances was compiled from various historical project records and studies, and a Monte Carlo simulation was utilized for synthetic variation generation to account for uncertainties. Advanced machine learning models were trained for cost and schedule risk prediction. The explainability of models was guaranteed by SHAP values. An interactive web-based virtual reality-enabled decision support dashboard for the selection of risk scenarios was developed, including labor shortages, rising prices of materials, and monsoon delays.

Results from the case building illustrate strong model performance, while XGBoost achieved R² values of 0.96 for cost and 0.94 for schedule prediction. The ML-optimized scenario provided 7.8% savings in cost and 27 days in schedule compression compared to the traditional baseline of Primavera P6. With the help of the interactive dashboard, project managers can evaluate different "what if" scenarios, which will facilitate decision-making based on predictions made by ML models.

This research provides a sustainable framework for BIM, machine learning, and interactive visualization in construction management for urban regions. The presented approach is very promising for further implementation in countries with similar construction project uncertainties.