Scholarlink

<p>Construction project management is characterized by inherent uncertainties, particularly in cost, schedule, and quality performance, which frequently lead to project underperformance or failure. Traditional risk assessment methods often lack the adaptability and predictive accuracy required to handle the multidimensional and data-intensive nature of modern infrastructure projects. To bridge the gap, this research crafts a data-driven model utilizing ensemble machine learning classifiers, such as Random Forest, XGBoost, and LightGBM, optimized using six metaheuristics: Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Ant Colony Optimization (ACO), Firefly Algorithm (FA), Grey Wolf Optimizer (GWO), and Black Widow Optimization (BWO). Using a comprehensive dataset sourced from civil projects executed during the period 2015–2020, the models were calibrated to forecast project risk status considering integrated cost, schedule, and quality indicators. The results indicated the LightGBM model optimized using PSO to yield the best prediction capability with an ROC–AUC of 0.92 and an F<sub>1</sub>-score of 0.88. Cost Variance Index (CVI) and Schedule Performance Index (SPI) were ranked as strongest predictors and validated their prospects as early indicators of construction project risk analytics. The work offers a scalable and explainable approach to proactive identification of risk and enables construction managers to implement focused plans of mitigation in highly sophisticated and digitally transforming environments. In itself an interweaving of ensemble learning and metaheuristics, the work demonstrates a significant advancement in predictive models of assessing risk and enables better-informed decisions in the management of infrastructure projects.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Ensemble machine learning for risk assessment in construction project management: applications in infrastructure development

  • 期刊论文

  • 发表日期：

    2025年8月21日

查看全文
  • Rakan Al Mnaseer

摘要

Construction project management is characterized by inherent uncertainties, particularly in cost, schedule, and quality performance, which frequently lead to project underperformance or failure. Traditional risk assessment methods often lack the adaptability and predictive accuracy required to handle the multidimensional and data-intensive nature of modern infrastructure projects. To bridge the gap, this research crafts a data-driven model utilizing ensemble machine learning classifiers, such as Random Forest, XGBoost, and LightGBM, optimized using six metaheuristics: Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Ant Colony Optimization (ACO), Firefly Algorithm (FA), Grey Wolf Optimizer (GWO), and Black Widow Optimization (BWO). Using a comprehensive dataset sourced from civil projects executed during the period 2015–2020, the models were calibrated to forecast project risk status considering integrated cost, schedule, and quality indicators. The results indicated the LightGBM model optimized using PSO to yield the best prediction capability with an ROC–AUC of 0.92 and an F1-score of 0.88. Cost Variance Index (CVI) and Schedule Performance Index (SPI) were ranked as strongest predictors and validated their prospects as early indicators of construction project risk analytics. The work offers a scalable and explainable approach to proactive identification of risk and enables construction managers to implement focused plans of mitigation in highly sophisticated and digitally transforming environments. In itself an interweaving of ensemble learning and metaheuristics, the work demonstrates a significant advancement in predictive models of assessing risk and enables better-informed decisions in the management of infrastructure projects.