Estimating temporal truck traffic patterns on rural highway systems: A spatially-aware machine learning approach

System- or network-wide truck traffic statistics are crucial for transportation planning, infrastructure design, and managing transport networks. However, producing these statistics is more challenging than for total traffic. Current methods depend largely on engineering judgment, making them labor-intensive, susceptible to human error, and inconsistent across jurisdictions. This study aims to address these challenges by proposing a machine learning (ML) solution, utilizing Random Forest spatial classification algorithm, to automate the assignment of short-duration count stations (SCSs) to temporal truck traffic pattern groups (TTPGs) and their attribution to road segments lacking volume data. The study further focuses on identifying the key factors that influence truck traffic patterns and creating models that address data limitations. The methodology was tested using data from Manitoba's 2019 traffic monitoring program, achieving an accuracy rate of over 80 %, showcasing its potential for broader use. The proposed ML approach offers an automated, reliable, accurate, and transferable method for analyzing truck traffic patterns, reducing processing time and reliance on subjective expertise. • Truck traffic pattern groups (TTPGS) are used to produce system-wide truck traffic statistics. • Short-duration count stations must be assigned to these groups manually. • Machine learning techniques can use spatial features to automate the assignment. • Key features, including spatial features, were identified for station classification. • Random forest classification was used to learn and classify datasets by TTPG.