Predicting average annual daily traffic using machine learning: a comparative study for infrastructure investment

The rapid growth in population and commercial activity necessitates substantial investment in transportation infrastructure. Given the constraints of limited budgets, accurate forecasting of traffic patterns is crucial for evaluating the benefits of such investments. This study addresses the need for precise traffic volume estimation by leveraging artificial intelligence techniques to enhance decision-making processes for infrastructure investments. We provide a comprehensive overview of traffic forecasting methodologies, emphasizing machine learning approaches and their application in various global contexts. Our analysis focuses on state roads in Türkiye, which handle approximately 70% of the country’s freight and passenger traffic. Using socioeconomic variables (population, GDP, and vehicle ownership) and infrastructure metrics (length of divided roads, bituminous hot mixture (BSK) paved roads, and surface coated roads) from provincial data, we developed traffic volume estimation models employing multiple linear regression, random forest regression, and support vector machine regression. The performance of these models was assessed using standard evaluation criteria. The results indicate that the random forest regression model outperformed the others in terms of accuracy. This study demonstrates that advanced machine learning techniques, particularly random forest regression, can significantly improve traffic forecasting precision. The insights gained are expected to guide the effective evaluation and prioritization of high-cost highway investments, thereby optimizing resource allocation and infrastructure development.