Comparative Performance Analysis of Gene Expression Programming and Linear Regression Models for IRI-Based Pavement Condition Index Prediction

Created by W.Langdon from gp-bibliography.bib Revision:1.8528

@Article{radwan:2025:Eng,

• author = "Mostafa M. Radwan and Majid Faissal Jassim and Samir A. B. Al-Jassim and Mahmoud M. Elnahla and Yasser A. S. Gamal",
• title = "Comparative Performance Analysis of Gene Expression Programming and Linear Regression Models for {IRI-Based} Pavement Condition Index Prediction",
• journal = "Eng",
• year = "2025",
• volume = "6",
• number = "8",
• pages = "Article No. 183",
• keywords = "genetic algorithms, genetic programming, gene expression programming",
• ISSN = "2673-4117",
• URL = "https://www.mdpi.com/2673-4117/6/8/183",
• DOI = "doi:10.3390/eng6080183",
• abstract = "Traditional Pavement Condition Index (PCI) assessments are highly resource-intensive, demanding substantial time and labor while generating significant carbon emissions through extensive field operations. To address these sustainability challenges, this research presents an innovative methodology using Gene Expression Programming (GEP) to determine PCI values based on International Roughness Index (IRI) measurements from Iraqi road networks, offering an environmentally conscious and resource-efficient approach to pavement management. The study incorporated 401 samples of IRI and PCI data through comprehensive visual inspection procedures. The developed GEP model exhibited exceptional predictive performance, with coefficient of determination (R2) values achieving 0.821 for training, 0.858 for validation, and 0.8233 overall, successfully accounting for approximately 82-85percent of PCI variance. Prediction accuracy remained robust with Mean Absolute Error (MAE) values of 12-13 units and Root Mean Square Error (RMSE) values of 11.209 and 11.00 for training and validation sets, respectively. The lower validation RMSE suggests effective generalisation without overfitting. Strong correlations between predicted and measured values exceeded 0.90, with acceptable relative absolute error values ranging from 0.403 to 0.387, confirming model effectiveness. Comparative analysis reveals GEP outperforms alternative regression methods in generalisation capacity, particularly in real-world applications. This sustainable methodology represents a cost-effective alternative to conventional PCI evaluation, significantly reducing environmental impact through decreased field operations, lower fuel consumption, and minimised traffic disruption. By streamlining pavement management while maintaining assessment reliability and accuracy, this approach supports environmentally responsible transportation systems and aligns contemporary sustainability goals in infrastructure management.",
• notes = "also known as \cite{eng6080183}",

}

Genetic Programming entries for Mostafa M Radwan Majid Faissal Jassim Samir A B Al-Jassim Mahmoud M Elnahla Yasser A S Gamal

Citations