Comparison of statistical and machine learning-based pavement deterioration models using PCI as a continuous variable

抄録

An accurate prediction of pavement condition is essential for effective pavement management systems, enabling road agencies to optimize maintenance strategies and extend the service life of road networks. The Pavement Condition Index (PCI) is one of the most commonly used pavement health indicators for assessing pavement deterioration. However, in many predictive models, PCI is often converted into an ordinal variable, which can result in the loss of valuable information regarding gradual deterioration patterns. This study aims to develop predictive pavement deterioration models treating PCI as a continuous variable to better capture the gradual deterioration of pavements over time. Both statistical and machine learning models were explored, including multiple regression, decision tree, random forest, and Artificial Neural Networks (ANN). Hyperparameter tuning was performed to optimize the performance of machine learning models, ensuring a balance between underfitting and overfitting. The models were evaluated based on R² and Root Mean Square Error (RMSE) values for both training and testing datasets, with performance parameters compared to identify the most effective model for predicting PCI as a continuous value. The findings indicate that random forest and double-layer ANN models possess superior predictive accuracy and generalization capabilities compared to other approaches, particularly in capturing complex deterioration patterns over time. The practical applicability of treating PCI as a continuous variable was explored, showing that it allows road agencies to identify critical deterioration stages and plan timely preventive maintenance and rehabilitation measures.