Analyzing the drivers of pedestrian activity at high spatial resolution

Pedestrian activity and mobility are key factors for transportation modeling, public health, and city planning. Traditionally, city agencies conduct manual counts by location and time as a proxy of pedestrian volume in surrounding area. With increasing data sources in the urban domain, particularly open city data, social media, wireless networks, new methods, and data analytics approaches are needed to extract deeper insights on pedestrian activity, volume, and mobility. This paper utilizes a dataset of pedestrian counts conducted by the New York City Department of Transportation (DOT) from 2007 to 2015, and integrates these data with geocoded public transit, land use, building, and streetscape information. Based on 100 locations defined by DOT, the study further explores correlations between pedestrian volume and the surrounding localized urban context. Through data mining, visualization, and statistical modeling, this study aims to (1) identify baseline and key indicators of pedestrian volume, and explore potential integration of new data sources with current methodologies; (2) use integrated approaches to exam how pedestrian volume changes by location and time; and (3) use regression models to predict pedestrian volume across other intersections in NYC. The results provide key indicators for pedestrian volume estimation without conducting manual counts. Overall, this study demonstrates a novel approach in utilizing new data sources and generating deeper analytical insights to understand the determinants and predictors of pedestrian activity.