Advanced Machine Learning Models for Predicting Project Performance in Complex Construction Environments | ||||
| MSA Engineering Journal | ||||
| Volume 4, Issue 1, June 2025, Page 1-14 PDF (417.98 K) | ||||
| Document Type: Original Article | ||||
| DOI: 10.21608/msaeng.2025.321348.1364 | ||||
 View on SCiNiTO | ||||
| Authors | ||||
Hyginus CO Unegbu  1; Danjuma S. YAWAS2
| ||||
| 1Ahmadu Bello University Nigeria | ||||
| 2Department of Mechanical Engineering Ahmadu Bello University Zaria Nigeria | ||||
| Abstract | ||||
| As the complexity of construction projects continues to grow, there is a critical need for more advanced predictive tools to forecast key performance outcomes, including cost overruns, schedule delays, quality, and safety risks. This study assesses the effectiveness of two machine learning models—Random Forests and Neural Networks—in predicting these outcomes using data from 1,000 construction projects. The performance of these models was compared with traditional methods, including the Critical Path Method (CPM) and Earned Value Management (EVM). The findings reveal that Random Forests consistently delivered the most accurate predictions, with the lowest Root Mean Squared Error (RMSE) for cost and schedule forecasting, as well as the highest precision and recall for identifying safety risks. Neural Networks also demonstrated significant improvements over traditional methods, though with slightly higher RMSE values. In contrast, CPM and EVM exhibited limited capabilities in handling the complex, multifactorial nature of modern construction projects, leading to higher prediction errors. These results suggest that adopting machine learning models can significantly enhance forecasting accuracy, decision-making, and risk mitigation in construction project management. Future research should apply these models to real-world projects to validate their operational effectiveness. | ||||
| Keywords | ||||
| Cost forecasting; construction management; machine learning; project performance | ||||
|
Statistics Article View: 30 PDF Download: 56 |
||||
