AI-Powered Forecasting of Environmental Impacts in Road Infrastructure Projects

Strategic Insights -

A new study has been conducted on the topic of environmental impact and construction costs in road infrastructure projects. Researchers from the Korea Institute of Civil Engineering and Building Technology have proposed a machine learning-based predictive framework using artificial neural networks and deep neural networks to accurately estimate these costs. The study has demonstrated promising results, with the deep neural network achieving average error rates of 27.1% for environmental load and 17.0% for construction cost at the planning stage.

Key Takeaways:

  • The researchers compiled a structured dataset of 150 completed national road projects in South Korea, covering both planning and design phases, to train and test their predictive framework.
  • The dataset focused on 19 high-impact sub-work types to reduce noise and improve prediction precision, and a hybrid imputation approach was applied to handle 4.47% missing data in the design-phase inputs.
  • The optimal artificial neural network (ANN) model achieved consistent performance across training and validation sets without overfitting, with a mean squared error (MSE) of 0.06 and root mean squared error (RMSE) of 0.24.
  • The deep neural network (DNN) model further improved performance, achieving average error rates of 27.1% (EL) and 17.0% (CC) at the planning stage and 24.0% (EL) and 14.6% (CC) at the design stage.
  • The synergy between deep learning and autoencoder-based feature selection offers a scalable and data-informed approach for enhancing early-stage environmental and economic assessments in road infrastructure planning.
  • The study supports the use of AI-powered forecasting in highway projects to improve project management and reduce environmental impacts.

Statistics:

  • 150 completed national road projects in South Korea were used to compile the dataset.
  • 19 high-impact sub-work types were focused on to reduce noise and improve prediction precision.
  • 4.47% of the design-phase inputs had missing data, which was handled using a hybrid imputation approach.
  • The optimal ANN model achieved MSE of 0.06 and RMSE of 0.24.
  • The DNN model achieved average error rates of 27.1% (EL) and 17.0% (CC) at the planning stage and 24.0% (EL) and 14.6% (CC) at the design stage.

Sources:

  • AI-Powered Forecasting of Environmental Impacts and Construction Costs to Enhance Project Management in Highway Projects (Buildings, 2025, 15(14): 2546)
  • Korea Institute of Civil Engineering and Building Technology, Department of Highway and Transportation Research (contact: Joon-Soo Kim)
  • MDPI AG (publisher of Buildings journal)
  • NewsRx LLC (copyright holder)