Conditional Density Plot Component Example on the Titanic Dataset

This simple workflow provides an example on how to use the verified component “Conditional Density Plot”. A general question it helps to answer when trying to find a good predictor for Y from feature X:
"Is X a good feature?"
A part of this question can be answered by answering instead:
"Does the distribution of X change by partitioning with class Y?".
Because if it does, X might be good predictor for Y. If it does too well (i.e. 100%), data leakage might be present.

Examples analysis is done on the titanic dataset. Each row of the data stands for a passenger, with columns like name, sex, age, booked class, fare, ... and whether he or she survived the incident. The task is to identify a feature that can act as a good predictor whether someone survived.
Drag and drop the component in another workflow from its KNIME Hub page (link below) to visualize a different dataset. Apply your custom settings by opening the component dialogue. More information is available in the component description.

Note that a Conditional Box Plot can be used for similar, less detailed analysis. The Conditional Density Plot offers more insight on the distribution itself.

External resources

• Conditional Density Plot component - KNIME Hub
• KNIME Verified Components - knime.com