Evaluating the relationship between walking and street characteristics based on big data and machine learning analysis

Abstract

The relationship between walking and the built environment is gaining increased attention for promoting sustainable transport and healthy communities. However, while pedestrians engage with the street environment, walkability assessments often overlook human-scale characteristics, focusing mainly on the neighborhood-level. Furthermore, traditional studies on walkability rely on limited and time-bound methods. To address these research gaps and obtain insights into the connection between walking and the built environment, this study utilizes machine learning techniques to scrutinize mobile-app data on pedestrian traffic alongside street characteristics. Tree-based algorithms are deployed to identify the association between walking volume and built environment features at the street-level, spanning distinct time periods. The pedestrian traffic data was gathered in Tel Aviv, Israel, while accounting for seasonal variations, weekdays, and time of day. Examining 20 street-level characteristics across 8000 segments furnishes new insights into the relative significance of various characteristics for walking, as well as street profiles linked to greater vs. lesser pedestrian activity. Notably, time variables emerge as crucial, with street features varying in importance across different time definitions. The study offers implications for decision-makers and urban planners by informing them of pedestrians' behaviors and preferences at the street-level, facilitating more efficient infrastructure investments and supporting planning decisions.