Prediction of International Roughness Index of Flexible Pavement Using Machine Learning-Based Predictive Framework in Ekiti State

• Oluwayinka. G. AKINWAMIDE

  Author

• Olugbenga O. AMU

  Author

• Christopher FAPOHUNDA

  Author

Accurate prediction of pavement roughness is essential for effective road design, maintenance planning, and long-term serviceability. The International Roughness Index (IRI) is a key indicator of ride quality, yet direct measurement can be resource-intensive. This study develops predictive models for IRI using commonly measured pavement indices: Present Serviceability Index (PSI), Pavement Condition Index (PCI), and Mean Texture Depth (MTD). Five modelling approaches were employed: Linear Regression, Random Forest (RF), Support Vector Machine (SVM), Gradient Boosting Machine (GBM), and Artificial Neural Network (ANN), applied to 480 highway sections in Ekiti State, Nigeria. Comparative evaluation using R², Root Mean Square Error (RMSE), and Mean Absolute Error (MAE) showed that all models provided reasonable predictive capability, with R² values ranging from 0.71 to 0.94 and RMSE values between 0.41 and 0.60. Ensemble methods GBM (R² ≈ 0.94, RMSE = 0.41) and RF (R² ≈ 0.92, RMSE = 0.45) consistently outperformed other models, effectively capturing nonlinear interactions among pavement indices. ANN and SVM offered moderate improvements over Linear Regression but were less accurate than ensemble methods. The findings highlight the applicability of machine learning for translating pavement condition indices into reliable IRI predictions, enabling data-driven decision-making. Integrating GBM and RF models into routine pavement evaluation frameworks can support timely maintenance interventions, optimize resource allocation, and improve road safety and ride quality. The study recommends regular model calibration with local pavement data to maintain accuracy and reinforce predictive reliability. Overall, ensemble learning approaches provide robust, cost-effective solutions for pavement roughness forecasting, demonstrating their potential to enhance sustainable infrastructure management in resource-constrained environments.

• FJET_21_40_40

Copyright (c) 2026 FUDMA Journal of Engineering and Technology

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.