This project consist of the definition of a time series ARIMAX model from NYC taxi fares data to forecast 100 instants in the future.
Caution
The taxi data .parquet files stored in the data directory are considerably large. The process of cloning the repository may take some time.
The data has been grouped by day in order to generate a time series index. The target variable is the taxi fare_amount, and the exogenous variable is the passenger_count. Initial analysis of the distribution through time shows inhomogenous variance of the data, and thus 1 differentiation step is applied to the model before optimising the autorregresive and moving average orders.
The prediction is made using and optimised ARIMAX(0, 1, 2) model.
In order to automate the setup of the project (virtual environment generation, activation, pip upgrade and required packages installation), both a .bat and a .sh file have been provided. Thus, to setup the project, run
path/to/project/setup.bat
in Windows based systems or
path/to/project/setup.sh
in Linux based systems.