Smart Mapping is an AI-driven, proprietary Virtualitics routine that helps you understand your data more quickly. To use Smart Mapping, you first select a target feature you want to understand in more detail, then input any features that you think might be related to that feature. Our algorithm will then:
- Create 1, 2 or 3 suggested mappings
- Rank the input features according to how relevant they are to the target feature
How to Access Smart Mapping
To access Smart Mapping, click on the icon on the toolbar. Drag the feature you want to understand into the Target field, then drag the features you want to use in the Features field. You can also click on Input All to include all the features. Hovering on a feature and clicking on the X will let you remove it from the list, and you can remove all features with many missing values by clicking on the “Remove sparse features” button next to Input All. Smart Mapping works with both categorical and numerical features but does not allow features where all the values are the same.
Each run of the routine produces a list of suggested mappings. The mappings consider the relative importance of the features and any correlations existing between them. Click on the Apply button to plot the associated mapping. If the target is a numerical feature it will be applied to the color dimension and split into quartiles. You can change the bin options by accessing the Color Settings in the Mapping panel. If a categorical feature is within the first 8 ranked features, it will be applied to Playback. The top binary categorical feature will also be mapped to Shape.
Feature Ranking and Correlations
The algorithm ranks features by relevance to the target. The rankings are scaled to 100 – the most important feature gets a score of 100 and all the others are scaled accordingly.
Through the Ranking dropdown menu, you can change the sort style to Correlation mode. In this mode, features will be sorted into correlated groups. Underneath the groups, there will be a list of features that are not correlated with any others.