The Evolution of Classical and Soft Computing Methods in Predicting Road Maintenance and Repair Costs: Approaches in the Literature and Future Perspectives

Road infrastructure is critical to the economic, social, and environmental sustainability of modern societies. This study compares classical methods (Multiple Linear Regression, Ridge, and LASSO) with soft computing techniques (Artificial Neural Networks, Fuzzy Logic, Random Forests, Gradient Boosting, Support Vector Machines, and Genetic Algorithms) for predicting road maintenance and repair costs. A comprehensive search has been conducted in Web of Science, and Scopus for studies published between January 2010 and March 2024. Boolean operators and specific key terms such as “road maintenance costs,” “soft computing,” and “classical prediction methods” have been used. The approach has been PRISMA-inspired but adapted for narrative review purposes; hence, no formal quality assessment or meta-analysis has been performed. Peer-reviewed journal articles have been included, while grey literature has been excluded to ensure methodological consistency. While classical methods offer simplicity and computational efficiency, they often fall short in addressing complex data structures such as non-linear relationships and multicollinearity. Conversely, soft computing techniques excel in modelling non-linear systems and managing uncertainties. Hybrid models combining classical and soft computing approaches enhance prediction accuracy by 20–30%, providing improved capabilities in modelling environmental factors. However, further research is required to evaluate their long-term performance and adaptability to diverse geographical conditions. This study highlights the theoretical advantages of hybrid models while offering practical solutions for sustainable infrastructure management. The findings provide policymakers and engineers with actionable insights, promoting efficient public resource use and sustainable development goals. Future research should focus on integrating IoT and big data analytics to address dynamic environmental variables, fostering innovation in infrastructure management.