Beyond ROC-AUC and KS: The Gini Coefficient, Explained Simply

The Gini Coefficient is a classification metric that measures how much better a model is at ranking positives higher than negatives compared to random guessing. Its calculation involves sorting data by predicted probabilities and plotting a Lorenz curve of cumulative population versus cumulative positives. The area under this Lorenz curve is then compared to the areas for a random model and a perfect model to derive the final Gini score. This score is standardized between 0 (random) and 1 (perfect), and is directly related to the ROC-AUC score through the formula Gini = (2 * AUC) - 1.