Scholarlink

<p>Timely completion of construction projects is vital for cost control, stakeholder satisfaction, and project success. In India, however, schedule adherence is frequently challenged by factors such as resource limitations, weather disruptions, and design changes. This study presents an AI-driven predictive modeling framework to forecast schedule performance using real-world data from Indian construction projects. Five machine learning models—multiple linear regression (MLR), support vector regression (SVR), artificial neural networks (ANN), random forest regressor (RFR), and gradient boosting machine (GBM)—were developed and evaluated. Models were categorized as white box, grey box, or black box based on interpretability. MLR, while transparent, showed limited performance in complex scenarios. SVR and RFR offered a balance between accuracy and explainability, whereas ANN and GBM achieved the highest predictive accuracy. GBM emerged as the top performer (R² = 0.93). Feature importance and sensitivity analyses identified planned duration, workforce size, equipment utilization, and material availability as key drivers of schedule adherence. The findings provide actionable insights for selecting suitable AI models in construction planning and underscore the value of integrating predictive analytics into schedule risk management for Indian 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

Exploring artificial intelligence applications in construction using a black grey white box approach for predicting project schedule performance in India

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月4日

查看全文
  • Ajay Tiwari,
  • Aslam Hussain

摘要

Timely completion of construction projects is vital for cost control, stakeholder satisfaction, and project success. In India, however, schedule adherence is frequently challenged by factors such as resource limitations, weather disruptions, and design changes. This study presents an AI-driven predictive modeling framework to forecast schedule performance using real-world data from Indian construction projects. Five machine learning models—multiple linear regression (MLR), support vector regression (SVR), artificial neural networks (ANN), random forest regressor (RFR), and gradient boosting machine (GBM)—were developed and evaluated. Models were categorized as white box, grey box, or black box based on interpretability. MLR, while transparent, showed limited performance in complex scenarios. SVR and RFR offered a balance between accuracy and explainability, whereas ANN and GBM achieved the highest predictive accuracy. GBM emerged as the top performer (R² = 0.93). Feature importance and sensitivity analyses identified planned duration, workforce size, equipment utilization, and material availability as key drivers of schedule adherence. The findings provide actionable insights for selecting suitable AI models in construction planning and underscore the value of integrating predictive analytics into schedule risk management for Indian infrastructure projects.