Feature importances

Marketing Analytics: Predicting Customer Churn in Python

Mark Peterson

Director of Data Science, Infoblox

Feature importances

Scores representing how much each feature contributes to a prediction
Effective way to communicate results to stakeholders

Which features are important drivers of churn?
Which features can be removed from the model?

Interpretability vs accuracy

Different models have different strengths
Need to balance prediction accuracy vs. interpretability

Random forest feature importances

random_forest = RandomForestClassifier()

random_forest.fit(X_train, y_train)

random_forest.feature_importances_

array([0.02716028 0.02960326 0.13820047 0.0700711  0.03798189 0.04722036
 0.13258838 0.07066312 0.02333901 0.02982636 0.13711014 0.0300661
 0.06351097 0.02860241 0.0382932  0.05200864 0.04375429])

Let's practice!

Marketing Analytics: Predicting Customer Churn in Python