Variable importance measures the contribution of each feature to a model's predictions, for example the cumulative impurity decrease for trees. It helps interpret and simplify the model, but tree-based importance can be biased in favor of variables with many values or correlated ones.
Example
modele.feature_importances_