Platt scaling

(machine learning) A technique for transforming the outputs of a classification model into a probability distribution over classes, based on a logistic regression model.