Multi-Classification Machine Learning Algorithm for Predicting Pavement Maintenance Treatments

Managing pavement assets poses a challenge for many highways worldwide, primarily attributed to funding restrictions and the absence of organized arrangement tools for maintenance and rehabilitation (M&R) activities to preserve road pavement assets. While pavement management systems (PMS) aid in predictions, the decision of when and how to administer treatments ultimately lies with pavement engineers. A model was developed to predict pavement treatment types with four input parameters: density, severity, average daily traffic (ADT) and road functional classification. The output parameters of the model correspond to various treatment techniques. The construction of this predictive model involved utilizing multi-classification machine learning (ML) algorithms, including support vector machine (SVM), Naive Bayes, K-Nearest Neighbour, and Random Forests. Furthermore, the model highlights the data categories that exert the most significant impact on the predictions. The accuracy of the model is noteworthy, with 81% superior for predicting untreated segments and exceeding 72% for predicting treated segments.