Saturday 4:45 PM–5:30 PM in Room #1025 (1st Floor)

Predicting Usage for Capital Bikeshare stations based upon Spatial Characteristics

Darshan Pandit

Audience level:
Intermediate

Description

A step after trying to model Continuous variables from Regression models, lies a very rewarding problem of estimating decisions which more lie in the discrete domain. In this talk, we work towards developing Logistic models to predict traffic for Capital Bikeshare, and work towards finding optimal station locations for Network expansion.

Abstract

  • We start ground up, by collecting and cleaning data from multiple data sources (Capital Bikeshare, Openstreetmap).
  • We then proceed towards developing a Linear regression model to predict the usage based upon spatial attributes of the stations, and critique its performance.
  • Using Biogeme, we then develop Logistic Models to estimate the same and compare their performance to its regression counter-parts.

The aim of this talk is to provide a primer to Logistic models and explore the effects of Spatial Characteristics on a Utility usage.

Sponsors

Become a sponsor.