Classification of roughness for asphalt pavement resurfacing treatments based on SVM-KNN machine learning algorithm using LTPP data

Abstract

Two supervised machine learning models, the Support Vector Machine (SVM) and SVM integrated with K-Nearest Neighbor (KNN), are adopted to classify pavement roughness levels of asphalt resurfacing treatments. Pavement related data are collected from the Long Term Pavement Performance (LTPP) database. Independent variables include overlay thickness, milling depth, using Recycled Asphalt Pavement (RAP), structural number, traffic level, annual precipitation, freeze index, service time and pre-treatment roughness. SVM-KNN result shows that pre-treatment roughness, overlay thickness, milling depth, service time, freeze index, traffic level all contribute to the classification. Using RAP only has slight influence on the post-treatment roughness. Incorporating KNN could improve the classification of data points close to the hyperplane. According to the machine learning performance measures including precision, recall, accuracy, and F1 score, the SVM-KNN obtained better classification results than SVM.