Application of artificial intelligence and machine learning in construction project management: a comparative study of predictive models

Abstract

This study examined the application of artificial intelligence and machine learning techniques in managing construction projects, focusing on planning, cost management, scheduling, quality control, and risk evaluation. This study employed Artificial Neural Networks (ANN), Support Vector Machines (SVM), and Random Forest (RF) algorithms to create predictive models. Findings revealed efficient resource allocation, with a minimal cost difference of 0.12% between projected and actual expenses. The Schedule Performance Index (SPI) of 1.04 suggested that the project was ahead of schedule, while a Cost Performance Index (CPI) of 0.91 indicated slight budget excesses. Quality measurements showed a defect rate of 2.5%, with three defects per 100 units. Among the tested ML models, Random Forest exhibited the best performance with an R² of 0.88, MSE of 1800, MEA of 36.25, and AUC of 0.95, outperforming ANN (R² = 0.85, MSE = 2000, MEA = 38.50, and AUC = 0.92) and SVM (R² = 0.80, MSE = 2500, MEA = 42.75, and AUC = 0.89). The safety performance index achieved 0.9 for compliance and 0.8 for training. These results show AI and ML can improve construction management, with RF being the top model for risk prediction and task management.

Graphical abstract