Urban bike flow prediction: traditional methods vs. explainable machine learning

• Grant number: 22/06328-7
• Support Opportunities: Scholarships abroad - Research Internship - Scientific Initiation
• Start date: August 23, 2022
• End date: December 22, 2022
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computer Systems
• Principal Investigator: Fabio Kon
• Grantee: Eduardo Bobrow Falbel
• Supervisor: Kay Axhausen
• Host Institution: Instituto de Matemática e Estatística (IME). Universidade de São Paulo (USP). São Paulo , SP, Brazil
• Institution abroad: Swiss Federal Institute of Technology Zurich, Switzerland
• Associated to the scholarship: 21/09285-4 - Urban cycling flow prediction using machine learning, BP.IC

Abstract

The push towards bicycle use is cities is a part of public policies which aim to promote the use of non-polluting means of transport that contribute to the population's overall health by inhibiting sedentary lifestyles. Deciding where to build cycling infrastructure is crucial to the viability of such policies. In this project we will evaluate data on bicycle trips in cities with public bike-sharing systems (BSSs). We shall utilize machine learning models in order to infer flows from socioeconomic, Point of Interest (POI) and topology (among other kinds of) data. Our goal is to utilize transfer-learning techniques in order to fit these models for use in systems without already established BSSs and, subsequently, allow for better planning regarding the placement of the stations for such systems and the construction of new cycling infrastructure. (AU)