Gini Impurity

A node-purity measure used to choose decision-tree splits: Gini(S)=1−∑kpk2\text{Gini}(S) = 1 - \sum_k p_k^2, where pkp_k is the share of class kk in the node. It is 0 for a pure node and 0.5 at maximum impurity for two classes.

When to use

When asked how a Decision Tree picks a split: it chooses the feature and threshold that minimise the weighted Gini of the child nodes.

See DS Lec 3.

Appears in