$$\text{gini impurity} = 1 - prob_{correct}^2 - prob_{incorrect}^2$$ The gini impurity indicates the likelihood of new data being misclassified, based off the distribution of labels of the training dataset.

Example:

flowchart TD A[Loves popcorn] A --> |True| B[Loves the song\nCorrect predictions: 80\nIncorrect predictions: 20\nGini impurity: 0.32] A --> |False| C[Deos not love the song\nCorrect predictions: 60\nIncorrect predictions: 40\nGini impurity: 0.48]

References⌗