Document Type : Research Paper
Authors
1
GIS Department, School of Surveying and Geospatial Engineering, College of Engineering, University of Tehran, Tehran
2
Department of GIS, School of Surveying and Geospatial Information, College of Engineering, University of Tehran, Tehran, Iran
3
Department of Geodesy, School of Surveying and Geospatial Engineering, College of Engineering, University of Tehran, Tehran, Iran
10.22059/jtcp.2024.370374.670427
Abstract
A significant portion of daily urban intra-city trips is aimed at accessing services, amenities, and goods that are not readily available in a specific area. Therefore, analyzing the frequently used trajectory and identifying the reasons for high traffic volumes on these trajectories can lead to a more accurate distribution of facilities, services, and proper land use allocation with the goal of reducing the number, distance, and time of intra-city trips. With the advent of Global Navigation Satellite Systems(GNSS) positioning sensors on smart phones, the real-time collection of individuals' positions, speed, acceleration, and more has become possible. Consequently, this research has sought to examine the possibility of using GNSS data recorded by smart phones to identify the transportation mode used by the user through four supervised machine learning models named Random Forest (RF), Gradient Boosting (GB), eXtreme Gradient Boosting (XGB), and Light Gradient Boosting Model (LightGM). For this purpose, two datasets, Microsoft Geolife and MTL 2017, which possess the necessary features for this goal, have been used as the input data. After extracting the features of each trajectory from these two datasets, with the aim of improving the models' performance and reducing processing time, among the available features, the most important ones have been identified and classification has been applied based on them. Among the models used LightGM and XGB models achieved the best performance for the first and second dataset with respective F1-Scores of 92.57% and 92.67% for test data. Out of a total of 1349 trips, this algorithm accurately estimated 1250 trips, contributing to sustainable urban transportation.
Keywords
Main Subjects
)