Department of Civil Engineering
Faculty of Engineering and Architecture Science
Using Wi-Fi data to infer transportation mode
Presenter: Arash Kalatian, Ryerson University
Date: Thursday March 22, 1:00PM—2:00PM
The intrinsic problems of conventional data collection methods have led researchers to explore the automated alternative approaches for acquiring data. One of the most prevailing methods of such approaches is the smartphone based data collection. In this study, we have utilized Wi-Fi communications obtained from smartphones of participants to predict their mode of transportation, i.e. walking, biking and driving. We deployed Wi-Fi sensors on four strategic locations in a closed loop of urban streets. These sensors are capable of recording and saving signal data of Wi-Fi enabled smartphones within their range, with no need for participant to log on. Participants were asked to walk, bike or drive in the area with their Wi-Fi enabled smartphones in their pockets. A dataset of 400 observations is created with 15 features, which are based on travel time, signal strength and connection time. The collected features were then used to classify participants based on their mode of transport. Three decision tree-based classifiers were trained: Decision Tree, Bagged Decision Tree and Random Forest, and more advanced methods are to be trained in the future. Validation results revealed that using the Random Forest method, transportation mode was predicted correctly in more than 80% of cases, with 87.72% accuracy for walking, 77.08% for biking and 83.64% for driving mode.
Arash Kalatian is a PhD student in transportation engineering at Ryerson university. Working as part of LITrans team under supervision of Dr. Bilal Farooq, his research mainly involves replacing conventional tools used in transportation engineering with modern alternatives. Currently, he is working on analyzing the interactions between pedestrians and autonomous vehicles, using a virtual reality environment. Besides that, he is also developing a framework to utilize Wi-Fi data for inferring traffic parameters, specifically mode of transportation.