Visualizing seasonal trends

We are going to do a new type of prediction involving the time variable. Before we build our model, we want to see whether the data is useful for making predictions. Our data must be such that present values have a relationship with past values. This characteristic is known as autocorrelation.

To make predictions, we need the following elements:

• Time-series data
• A trend line from the data linear regression
• Errors of the data linear regression

We will use an 8-year air passenger sales dataset to do a time-series analysis. In the following screenshot, look at the plot of the aforementioned elements. We can see that the data is probably useful enough for us to do a 1-year forecast. We will look at a prediction for Year 9 and see whether the Year 9 forecast makes sense based on past behavior and our data experience. Remember that everything can be validated, but ultimately, we have to judge whether the results make sense or not using...