Spatio-temporal Modeling of Yellow Taxi Demands in New York City Using Generalized STAR Models

TL;DR

This paper introduces a generalized spatio-temporal autoregressive model for predicting taxi demand in New York City, effectively capturing dynamic spatial-temporal variations and outperforming traditional models like VAR.

Contribution

The study develops a novel generalized STAR model with LASSO penalization for high-dimensional data, improving prediction accuracy and interpretability for urban taxi demand forecasting.

Findings

Proposed models outperform VAR in MSPE.

Model is computationally efficient for real-time use.

Framework is easily interpretable and practical for taxi operators.

Abstract

A highly dynamic urban space in a metropolis such as New York City, the spatio-temporal variation in demand for transportation, particularly taxis, is impacted by various factors such as commuting, weather, road work and closures, disruption in transit services, etc. To understand the user demand for taxis through space and time, a generalized spatio-temporal autoregressive (STAR) model is proposed in this study. In order to deal with the high dimensionality of the model, LASSO-type penalized methods are proposed to tackle the parameter estimation. The forecasting performance of the proposed models is measured using the out-of-sample mean squared prediction error (MSPE), and it is found that the proposed models outperform other alternative models such as vector autoregressive (VAR) models. The proposed modeling framework has an easily interpretable parameter structure and practical to be applied by taxi operators. Efficiency of the proposed model also helps in model estimation in real-time applications.