Using structural estimation methods from economics to model user behaviour in bike-sharing systems
Submitted by Ashish Kabra (@akabra) on Wednesday, 11 April 2018
The cities of Paris, London, Chicago, and New York (among many others) have set up largescale bike-share systems to facilitate the use of bicycles for urban commuting. This talk estimates the impact on bike-share ridership of two facets of system performance: accessibility (how far the user must walk to reach stations) and bike-availability (the likelihood of finding a bicycle). My analysis is based on a structural modelling pricinples from economics to allow causal inference and accurately model and estimate user behavior while using only aggregated station use data for estimation. The proposed model is able to incorporate the real time changes in bike-availability information, and include hyperlocal data on potential demand sources for accurate predictions. This model turns out to be computationally expensive; we transform our estimation from the time domain to the “local-stockout- state” domain to address this. The model has about 50% better prediction performance than common methods. I illustrate the use of these estimates in identifying neighborhoods and times to target for improvements, and in comparing alternate operational improvements and station networks.
- Introduction of bike-share context
- Challenge of estimating user preference from station level data
- Model formulation and estimation
- Computational Challenge - Solution
- Illustration of prescriptive power of method to solve system design problems
Nothing. The talk will be self-explanatory
Ashish Kabra is a tenure-track faculty member in the Department of Operations and Information Techonology at the University of Maryland, College Park. His expertise is in using developing and applying estimation algorithms to study new business models such as bike-share systems (eg: Citibike) and marketplaces (eg: Uber). He has studied topics related to “accessibility” (sufficient reach), availability (service is available when a user needs it), and that of effectiveness of promotions in scaling marketplaces. He has also studied online grocery retail models (eg: Amazon Fresh), specifically its financial and environmental concerns using mathematical economics models.
His research work has been published in Management Science and has been invited at several international conferences including INFORMS, MSOM, POMS. His research work has won the MSOM Best Student Paper Award, a runner up at POMS Best Student Paper Award in Sustainability, and a third place at IBM Best Student Paper Award in Service Science.
He did his graduate studies in Operations Management at INSEAD, France and undergraduate studies in Computer Science from BITS-Pilani, India.
He has also worked for Adobe Systems and a high-tech supply chain analytics startup in the past and consulted with data science and management teams at sharing economy startups.