Interpretation of contextual influences with explanatory tools: Travel mode likelihood mapping using GPS trajectories

Abstract

Past studies have failed to address the spatially and temporally varying impacts of environmental factors regarding the uncertain geographic context problem. This study seeks to provide an innovative framework to facilitate the understanding of spatially and temporally varying impacts of multiple contexts on individuals' travel modes using GIS and machine learning techniques. It adopts machine learning techniques to create likelihood maps to predict the spatiotemporal patterns of individual travel behaviors and uses explanatory tools to explore the spatially and temporally varying impacts. The most notable change at a local level in the spatial dimension was that assaults and offenses involving children turned out to be important in two selected communities in Chicago. Regarding the temporally varying impact, batteries, other offenses, and robberies showed negative associations with the walking prediction to some extent at the afternoon peak (5–7:59 p.m.) during weekdays. The proposed approach will enable meaningful interpretation of complex interactions between multiple environmental factors and individual travel behaviors to suggest policies in urban planning and design.