# Applying a GM (1, 1)-BPNN to predict pavement Rutting Depth Index in hot and humid region: A case study in Guangdong, China

## TL;DR

This paper proposes a new model to predict pavement rutting in hot and humid regions, using data from Guangdong, China, and identifies key factors affecting pavement performance.

## Contribution

A novel GM(1,1)-BPNN model is introduced for accurate prediction of pavement rutting depth index in hot and humid regions.

## Key findings

- The GM(1,1)-BPNN model outperformed baseline models with MAE of 0.068 and R2 of 0.79.
- SHAP analysis revealed that maintenance fund and pavement temperatures significantly impact rutting depth index.

## Abstract

Pavement performance prediction plays a crucial role in formulating scientific pavement maintenance plans. However, current research on how the rutting depth index (RDI) in hot and humid regions is affected by multiple influencing factors and the development of accurate prediction indicators remains insufficient. To establish a scientific basis for maintenance, the research team collected maintenance, traffic, pavement surface and internal temperature, climate, and road condition data from 2015 to 2021 for a freeway section located in Foshan, China, a typical hot and humid region. Then, a combined predictor, GM(1,1)-BPNN, was proposed to conduct accurate RDI prediction for the pavement. Furthermore, the SHapley Additive exPlanation (SHAP) method was employed to analyze the impact of each influencing factor on RDI in greater detail. The results indicated that 1) The proposed combined model has a higher prediction performance. Validated by validation set, the MAE, MSE, RMSE as well as R2 were 0.068, 0.004, 0.068, 0.79, respectively, surpassing the baseline models PPI and GM (1, 1); 2) The SHAP analysis shows that maintenance fund, middle layer maximum temperature, integrated radiation, and pavement surface maximum temperature have a more significant impact on RDI. The conclusions of the paper provide a theoretical basis for road administrations to formulate scientific maintenance plans and contribute to understanding the impact of climatic and traffic environments on RDI.