Data Mining Approach-Based Damage Identification for Asphalt Pavement Under Natural Disaster Conditions

Road performance can also decline due to natural disasters such as earthquakes, often in Indonesia. Given the high risk of natural disasters in Indonesia, it is important to consider their impact. Therefore, it is necessary to prepare for road rehabilitation and reconstruction quickly and accurately. This research aims to identify potential factors causing road damage by developing an approach to obtain predictions of road damage levels due to natural disasters by utilizing the availability of historical data, developing a decision support system to rehabilitate and reconstruct roads after disasters, and developing a road damage model due to earthquakes using data mining. The data was used to assess the condition of the national road pavement in Central Sulawesi and identified the disaster events as earthquakes that originated from the USGS. Data processing uses a data mining (DM) approach, which includes three models. The results found that the SVM modeling with the DM approach had a high accuracy rate of 0.91 ± 0.01, RMSE 0.70 ± 0.02, and MAD 0.42 ± 0.01. SVM achieves the highest accuracy after 20 runs. The best hyperparameters to accomplish a fit SVM model are ϵ = 0.07 ± 0.01 and γ = 0.05 ± 0.00. Meanwhile, for ANN, the hyperparameters are H = 3 ± 1. The earthquake’s magnitude (27%) and depth (24%) contribute to road damage.