Feature Selection by Conformal Predictor

Abstract

Abstract. In this work we consider the problem of feature selection in the context of conformal prediction. Unlike many conventional machine learning methods, conformal prediction allows to supply individual predictions with valid measure of confidence. The main idea is to use confidence measures as an indicator of usefulness of different features: we check how many features are enough to reach desirable average level of confidence. The method has been applied to abdominal pain data set. The results are discussed.

Learning by Conformal Predictors with Additional Information

“…To avoid the influence of redundant attributes, we use some selected symptoms here. For each disease group, we use 5 most relevant symptoms selected in [7] as "usual" attributes because these 5 selected symptoms could provide the similar confidence level as whole set of symptoms. The features provided by experts in [1] are used as privileged attributes.…”