Authors: Kangjae Lee*, Illinois Informatics Institute, University of Illinois at Urbana Champaign
Topics: Geographic Information Science and Systems, Geography and Urban Health, Human-Environment Geography
Keywords: predictive analytics, machine learning, GIS, potential mapping, GPS
Session Type: Paper
Start / End Time: 9:55 AM / 11:35 AM
Room: Balcony A, Marriott, Mezzanine Level
Presentation File: No File Uploaded
The impact of environmental contexts on physical activity, like walking and jogging showed inconsistent outcomes often against underpinning ecological models failing to provide reliable empirical evidence and make generalizable knowledge. Previous studies did not consider complex interactions between human behaviors and environments and a time context, which is an important aspect to understand neighborhood effect on human behaviors, and missed interpreting local impact of contextual influences regarding spatial non-stationarity. To deal with the suggested three viewpoints, this study seeks to provide an innovative approach to potential mapping of non-motorized transport, as a subset of physical activity, using GIS and machine learning techniques. Potential mapping allows exploring distinct patterns of walking and biking and investigating local impact of environmental factors on the two transport modes. Support vector machine, random forest, and gradient boosting classifiers are evaluated to select the best predictive model to create potential maps based on GPS trajectories of individuals. Visual exploratory analysis and the interpretation of the local impact on non-motorized transport modes are conducted on the generated potential maps in space and time contexts. This study will contribute to the introduction of the data-driven approach to public health and transport research beyond the traditional statistical approach, by utilizing machine learning techniques. In addition, the proposed approach to generate potential maps will advance GIS mapping by employing machine learning techniques to predict spatio-temporal patterns regarding public health using daily movement data of individuals.