Gaussian mixture pdf in one-class classification: computing and utilizing confidence values

In this study a confidence measure for probability density functions (pdfs) is presented. The measure can be used in one-class classification to select a pdf threshold for class inclusion. In addition, confidence information can be used to verify correctness of a decision in a multi-class case where for example the Bayesian decision rule reveals which class is the most probable. Additionally, using confidence values - which represent in which quantile of the probability mass a pdf value resides ([0,1]) - is often straightforward compared to using arbitrarily scaled pdf values. As the main contributions, use of confidence information in classification is described and a method for confidence estimation is presented

1 Introduction

One-class classification, also called as novelty detection, outlier detection, or data description ([7]), can be used to detect uncharacteristic observations. One-class classification is necessary when samples can be obtained only from a single known class, for example, normal operation mode in motor condition monitoring where all failure modes are not known. One-class classification is also useful when the background class contains enormous variations making its estimation unfeasible, for example, background class in object detection: the background class should contain everything except the object to be detected.