Bending the Curve: Improving the ROC Curve Through Error Redistribution

TL;DR

This paper introduces a meta-learning method that dynamically adjusts classification thresholds based on data difficulty features to improve the ROC curve, applicable to any classifier treated as a black box.

Contribution

It presents a novel post-processing algorithm for error redistribution that enhances ROC performance by leveraging difficulty-related features without modifying the base classifier.

Findings

Improved ROC curves on synthetic data

Enhanced true-positive/false-positive trade-off on real data

Applicable to any black-box classifier

Abstract

Classification performance is often not uniform over the data. Some areas in the input space are easier to classify than others. Features that hold information about the "difficulty" of the data may be non-discriminative and are therefore disregarded in the classification process. We propose a meta-learning approach where performance may be improved by post-processing. This improvement is done by establishing a dynamic threshold on the base-classifier results. Since the base-classifier is treated as a "black box" the method presented can be used on any state of the art classifier in order to try an improve its performance. We focus our attention on how to better control the true-positive/false-positive trade-off known as the ROC curve. We propose an algorithm for the derivation of optimal thresholds by redistributing the error depending on features that hold information about difficulty. We demonstrate the resulting benefit on both synthetic and real-life data.