Optimizing Road Infrastructure Financing through Machine Learning-Driven Cost Forecasting Models: Evidence from Indian Highway Projects.

Abstract

In the context of India’s ambitious infrastructure expansion agenda, road development projects often encounter significant financial inefficiencies due to inaccurate cost estimations, time overruns, and suboptimal resource allocations. This study employs advanced machine learning (ML) algorithms to optimize cost forecasting models tailored for Indian highway infrastructure. Leveraging a dataset comprising 150 highway projects executed under the National Highways Development Programme (NHDP), this research systematically compares the predictive efficacy of multiple ML models—namely Random Forest Regression (RFR), Gradient Boosted Decision Trees (GBDT), and Artificial Neural Networks (ANN)—in estimating total project costs. The results reveal that ML-enabled forecasting frameworks substantially outperform conventional linear regression approaches, offering enhanced accuracy and real-time adaptability. Furthermore, the study proposes a strategic framework for integrating ML-driven insights into public and private financing decisions, potentially reducing fiscal risk and improving capital allocation across infrastructure portfolios. The findings have significant implications for infrastructure economists, financial analysts, and public policy architects in emerging economies.