Explaining Walking in Cities – a Machine Learning Approach

Abstract

A large body of research has developed on walking and walkability, in part in response to increasing concerns over people’s health, climate change, livability, and social cohesion. Literature shows that some built environment and socio-demographic characteristics influence walking rates more than others.

Different approaches and methods have been used to study the relationship between the built environment characteristics, socio-demographic variables and walking patterns. Yet, so far very few studies have applied machine learning tools to study and explore these relationships. This research aims to start filling this void.

The study draws on a dataset contains details about trips made by over 37,000 respondents in the Tel-Aviv metropolitan area. The detailed data allow us to differentiate between walk-only trips and walk trips that are combined with other modes of transport. Our results show that the built environment shapes walk-only trips more than walking as an access or egress mode.