*Chapter 13*: Training, Validating, and Running the Model

In this chapter, we will apply the forecast time-series model to a real-life dataset to predict automobile sales in the US, using Kaggle retail sales data.

We have quarterly data for the years 2012 to 2019. We will design, train, and test the model and see whether it does a good job of making predictions.

In **Machine Learning** (**ML**), when working with statistical groups, linear regression, or time series, you have to apply your experience to do an initial quality check of data with a chart. In a time-series forecast, you use your judgment to see whether the data has autocorrelation. That means that the past has influence over the present and is useful to predict the future using a forecast.

Many time-series datsets have two components that need prediction – a season component and a growing decreasing trend. The season component is when data has cycling peaks depending on a year's seasons.

After these calculations...