Taxi and Mobility: Modeling Taxi Demand Using ARMA and Linear Regression.

Demand for taxis varies over time. Finding the factors that affect demand and understanding the dynamics of taxi demand will benefit different groups. It helps planners to improve their transportation systems and also help drivers to reduce their empty vacant time. This paper focuses on modeling and analyzing the taxi demand, using the demand for other transportation modes and weather conditions. We started with the linear regression model, and then combined with a time series models. This combined model is called "linear regression with ARMA errors". (ARMA: Auto Regressive Moving Average). We considered the borough of Manhattan in New York City, and collected three months of data of yellow cabs, subway, Uber, bike, temperature, and precipitation. The observations for each variable, are aggregated hourly. We analysed these data sets and removed their seasonality, and then applied the models to those data sets. The power of time series model in explaining a variable is discussed and the results show that the developed model outperforms the linear regression, by improving the R-squared more than %30. [ABSTRACT FROM AUTHOR]