Discussing traditional pipelines

The initial approach involves statistical modeling, using models such as ARIMA, ARIMA with exogenous variables (ARIMAX), and Auto ARIMA. To work with them, we need to address two additional challenges: ensuring the stationarity of the time series and determining the appropriate model order.

Statistical models perform better when applied to stationary time series. Traditional statistical time series models such as ARIMA are more effective when dealing with stationary time series. Resolving this issue will be part of the preprocessing phase.

The second challenge lies in the modeling phase, which involves understanding the dataset, determining the appropriate lags, and defining time windows. We will approach the solution manually using the Auto ARIMA algorithm, which handles hyperparameters automatically.

Preprocessing

Various functions can be employed to transform non-stationary time series data into a format suitable for our models. Examples...