In this recipe, we focus on using the ARIMA class models to forecast future observations of a given time series.
We compare the forecasting performance of the models we built in the Modeling time series with ARIMA class models recipe, where we investigated Google's stock prices in 2015-2018. We manually selected an ARIMA(2,1,1) model, while auto_arima suggested ARIMA(3,1,2). In this recipe, we use both models as they were initially estimated using different libraries, offering slightly different possibilities in terms of forecasting.
We forecast Google's weekly stock prices over the first 3 months of 2019.
