SARIMA vs ARIMA forecasts
In this exercise, you will see the effect of using a SARIMA model instead of an ARIMA model on your forecasts of seasonal time series.
Two models, an ARIMA(3,1,2) and a SARIMA(0,1,1)(1,1,1)\(_{12}\), have been fit to the Wisconsin employment time series. These were the best ARIMA model and the best SARIMA model available according to the AIC.
In the exercise you will use these two models to make dynamic future forecast for 25 months and plot these predictions alongside held out data for this period, wisconsin_test.
The fitted ARIMA results object and the fitted results object are available in your environment as arima_results and sarima_results.
This exercise is part of the course
ARIMA Models in Python
Exercise instructions
- Create a forecast object, called
arima_pred, for the ARIMA model to forecast the next 25 steps after the end of the training data. - Extract the forecast
.predicted_meanattribute fromarima_predand assign it toarima_mean. - Repeat the above two steps for the SARIMA model.
- Plot the SARIMA and ARIMA forecasts and the held out data
wisconsin_test.
Hands-on interactive exercise
Have a go at this exercise by completing this sample code.
# Create ARIMA mean forecast
arima_pred = arima_results.____
arima_mean = arima_pred.____
# Create SARIMA mean forecast
sarima_pred = sarima_results.____
sarima_mean = ____
# Plot mean ARIMA and SARIMA predictions and observed
plt.plot(dates, sarima_mean, label='SARIMA')
plt.plot(dates, arima_mean, label='ARIMA')
plt.plot(wisconsin_test, label='observed')
plt.legend()
plt.show()