Using Rattle to forecast the demand

In this section, we'll use Rattle for a quick correlation analysis and to create a model to forecast the bike demand.

Correlation Analysis with Rattle

Our dataset has three possible target variables: cnt, registered, and casual. Rattle doesn't handle multiple targets. We can create a model for registered and a second model for casual and add both values to have the total number of users, or we can build a model for cnt (the total amount). We will only create a model for cnt because we're interested in the level of activity and this variable will provide it.

Load the dataset into Rattle and set instant to Ident and dteday, registered and casual to Ignore, and set the rest of variables to Input.

To perform the correlation analysis, go to the Explore tab, select the Correlation radio button and, finally, press the Execute button, as shown in this screenshot:

Rattle will return us a correlation matrix and a correlation plot. The correlation matrix is a matrix that...