Unlocking insights from commercial vehicle data: A machine learning approach for predicting commercial vehicle classes using Michigan State data (1999–2017)

Abstract

Commercial vehicles have a significant economic effect by moving goods and services across national borders and around the globe. They are also responsible for most commercial transactions in several industries, including manufacturing, retail, agriculture, and building. Its characteristics, such as heavyweight and big vehicle size, also affect how traffic moves and streams behave. As a result, it adds to growing accident rates, congestion, pollution, and sidewalk deterioration. This paper uses the commercial vehicle survey (CVS) from Michigan state based on different establishments to investigate the pattern movements of commercial vehicles between 1999 and 2017. This study aims to develop predictive commercial vehicle classes through machine learning techniques. This study uses three machine learning methods to predict the Commercial Motor Vehicle (CMV) class (Naive Bayes, Linear SVM, and decision tree). A feature selection study selects the significant attributes for CMV class prediction. The accuracy of the classification prediction model methods was compared through the training and testing phases. The results show that The CVS was successfully used to classify commercial vehicles with accuracy greater than 89 %.