Forecasting univariate time series with auto_arima
For this recipe, you will need to install pmdarima, a Python library that includes auto_arima – a tool designed to automate the optimization and fitting of ARIMA models. The auto_arima implementation in Python is inspired by the popular auto.arima from the forecast package in R.
As you've seen in earlier recipes, determining the correct orders for the AR and MA components can be challenging. While techniques like examining ACF and PACF plots are helpful, finding the optimal model often involves training multiple models – a process known as hypeparameter tuning, which can be quite labor-intensive. This is where auto_arima shines as it simplifies the effort.
Instead of the naïve, brute force, approach of manually conducting a grid search to try every parameter combination, auto_arima uses a more efficient approach to finding the optimal parameters. The auto_arima function uses a stepwise algorithm that is faster...
