Integration of GIS and machine learning techniques to investigate the impact of environmental contexts on travel modes

Related to the promotion of physical activity, a growing body of research has adopted the definition of active travel modes. Active travel modes have made a great contribution to overall physical activity and, therefore, it is important to understand the active travels associated with environmental facilitators or barriers in physical activity and transportation research. Residential neighborhoods around individuals’ home locations were a primary focus in previous studies to examine the associations between active travels and environmental factors and, for the last decade, researchers have begun using global positioning system (GPS) trajectories of individuals to consider their daily paths for actual exposure estimation to various environments. Empirical findings in the existing studies, however, showed inconsistent outcomes of the associations. In addition, more advanced analytical approaches have not yet been explored, regardless of a large amount of GPS trajectories in hand, which have great potential to find more valuable and various outcomes. Thus, this study seeks to provide comprehensive data-driven approaches to further investigate the associations between travel modes and environmental contexts using the geographic information system (GIS) and machine learning techniques. An automatic travel mode classification algorithm is developed using GPS and accelerometer data to advance travel mode detection in health and transportation research. When it comes to exposure estimation to various environments, this study focuses on buffer analysis, which has been widely used in previous studies, and examines how distance, as one of the buffer characteristics, can affect findings of the associations between travel modes and environmental factors to give insights into accurate estimation of immediate surroundings along the daily trajectories of individuals. In addition, a novel framework is proposed and adopted to perform mapping of travel modes and explore complex contextual influences on travel modes at different levels of scales using machine learning models. In the era of big data, this dissertation suggests methodological directions for various fields of study to adequately deal with a large quantity of sensor data collected from many participants, derive informative measures for classifying health behaviors from the sensor data, and conduct exploratory analyses and produce meaningful knowledge using machine learning models with GIS data.U of I OnlyAuthor requested U of Illinois access only (OA after 2yrs) in Vireo ETD syste