Platt scaling

Platt scaling, also known as Platt calibration, is a way of transforming the outputs of a classification model into a probability distribution over classes. It works by fitting a logistic regression model to the scores that the classifier produces, so that the classifier's raw scores are mapped onto probabilities. The method was invented by John Platt in the context of support vector machines, where it replaced an earlier method by Vapnik. Although it was devised for support vector machines, it can be applied to other classification models as well, because it needs only the scores the model outputs.