Feature selection

The basic idea behind feature selection is to select features that show high importance for the target. In addition, we want to remove any features that are highly cross-correlated (or have high MI values) to other features. The selected set of features are represented as feature lists in DataRobot. If you click on the Feature Lists menu on the top left of the page, as shown in the following screenshot, you will see the feature lists that DataRobot has created for the dataset:

Figure 5.24 – Feature Lists

Here, you will see a list that contains all the raw features, ones that have selections based on univariate analysis (that is, analysis of features one at a time), and also ones that have the most important features. The DR Reduced Features M8 list or the Univariate Selections list look like good starting points. Click on the Project Data menu to go back to the data view. Now, let's inspect the univariate list by selecting Univariate...