To enhance the scientific rigor and precision of urban road maintenance decision-making, this study proposes an ensemble learning-based method for predicting the asphalt pavement condition index (IPCI). Based on historical pavement inspection and maintenance data for urban roads in Beijing from 2017 to 2024, a multidimensional feature dataset encompassing pavement usage indicators, performance indicators, and environmental impact indicators was constructed. Feature selection was performed using Pearson correlation coefficients and the LASSO method. Based on these selected features, 11 widely used machine learning models were trained, and their hyperparameters were optimized. After a comprehensive comparison of predictive accuracy, the five best-performing models, including XGBoost, LightGBM, random forest, gradient boosting tree, and CatBoost, were selected as base learners to construct an ensemble prediction model using weighted average voting regression. The experimental results demonstrate that the proposed ensemble model performs well on the test set, with R² = 0.888 5, ERMSE = 0.057, and EMAE = 0.043 1. The model exhibits an extremely narrow range of generalization errors, demonstrating its effectiveness in handling complex nonlinear relationships. SHAP-based interpretability analysis identifies pavement age, maintenance grade, annual average daily traffic, and alligator cracking as the key factors affecting IPCI. The study confirms that this ensemble prediction method can significantly improve the accuracy and interpretability of asphalt pavement condition prediction, providing theoretical support for formulating urban road maintenance specifications, optimizing the allocation of maintenance funds, and implementing preventive maintenance systems.
Editorial Department of Road Traffic Management. Development and changes of road traffic management in Beijing (Ⅵ): “past lives” of zebra crossing[J]. Road Traffic Management,2024(6): 84-85.
XuXizhong, WeiJincheng, YanXiangpeng, et al. Review and outlook for long-life asphalt pavement[J]. Journal of China & Foreign Highway, 2023, 43(1): 36-43.
[5]
AliA A, MiladA, HusseinA, et al. Predicting pavement condition index based on the utilization of machine learning techniques: A case study[J]. Journal of Road Engineering, 2023, 3(3): 266-278.
LiXiaohua, SongLiang, YeWei, et al. Distribution characteristics of internal distress and maintenance decision-making for asphalt pavement based on three-dimensional ground-penetrating radar[J]. Journal of China & Foreign Highway, 2025, 45(3): 221-231.
[8]
LeeJ H, JungD H, LeeM S, et al. A feasibility study for the prediction of concrete pavement condition index (CPCI) based on machine learning[J]. Applied Sciences, 2022, 12(17): 8731.
[9]
YuT, PeiL I, LiW, et al. Pavement surface condition index prediction based on random forest algorithm[J]. Journal of Highway and Transportation Research and Development (English Edition), 2021, 15(4): 1-11.
[10]
SongY, WangY D, HuX B, et al. An efficient and explainable ensemble learning model for asphalt pavement condition prediction based on LTPP dataset[J]. IEEE Transactions on Intelligent Transportation Systems, 2022, 23(11): 22084-22093.
ZhangLijuan, YangYuan, MeiCheng, et al. Comparing predictions of asphalt pavement performance between ARIMA and support vector machine models[J]. Journal of China & Foreign Highway, 2022, 42(2): 52-56.
[13]
ZhangR, HuP K, ZhongY H, et al. Research progress on evaluation and prediction of degradation in service performance for asphalt pavement[J]. Journal of Traffic and Transportation Engineering (English Edition), 2025, 12(4): 1011-1039.
ZhangPanpan, QuanLei, LiSili, et al. Model for asphalt pavement performance prediction based on LSTM-BPNN neural network[J]. Journal of Highway and Transportation Research and Development, 2023, 40(S2): 54-60.
[16]
MersM, YangZ Y, HsiehY A, et al. Recurrent neural networks for pavement performance forecasting: Review and model performance comparison[J]. Transportation Research Record: Journal of the Transportation Research Board, 2023, 2677(1): 610-624.
[17]
GuoR H, FuD L, SollazzoG. An ensemble learning model for asphalt pavement performance prediction based on gradient boosting decision tree[J]. International Journal of Pavement Engineering, 2022, 23(10): 3633-3646.
[18]
PeiL L, YuT, XuL, et al. Prediction of decay of pavement quality or performance index based on light gradient boost machine[M]//Advances in Intelligent Automation and Soft Computing. ChamSpringer International Publishing,2021: 1173-1179.
[19]
余婷. 基于集成学习的路面技术状况指数预测方法研究[D]. 西安: 长安大学, 2022.
[20]
YuTing. Ensemble learning based prediction method for pavement maintenance quality index[D]. Xi’an: Chang’an University, 2022.
PeiLili, YuTing, ZhangYanhong, et al. Prediction of national and provincial road pavement performance based on Bayesian-LightGBM model[J]. Journal of Highway and Transportation Research and Development, 2023, 40(11): 18-26.
[23]
ZhangX M, RenJ W, FengQ H, et al. Prediction of refracturing timing of horizontal wells in tight oil reservoirs based on an integrated learning algorithm[J]. Energies, 2021, 14(20): 6524.
[24]
WangY C, SongH M, WangJ S, et al. Multi-strategy fusion binary SHO guided by Pearson correlation coefficient for feature selection with cancer gene expression data[J]. Egyptian Informatics Journal, 2025, 29: 100639.
[25]
CuiJ W, CheA L, LiS, et al. A logistic-Lasso-regression-based seismic fragility analysis method for electrical equipment considering structural and seismic parameter uncertainty[J]. Earthquake Engineering and Engineering Vibration, 2025, 24(1): 169-186.
[26]
ZhangJ J, ChengT, GuoW, et al. Leaf area index estimation model for UAV image hyperspectral data based on wavelength variable selection and machine learning methods[J]. Plant Methods, 2021, 17: 49.
[27]
FeiS P, HassanM A, XiaoY G, et al. UAV-based multi-sensor data fusion and machine learning algorithm for yield prediction in wheat[J]. Precision Agriculture, 2023, 24(1): 187-212.