Asymmetric Error Control for Binary Classification in Medical Disease Diagnosis

Abstract

In binary classification applications, such as medical disease diagnosis, the cost of one type of error could greatly outweigh the other enabling the need of asymmetric error control. Due to this unique nature of the problem, where one error greatly outweighs the other, traditional machine learning techniques, even with much improved accuracy, may not be ideal as they do not provide a way to control the false negatives below a certain threshold. To address this need, a classification algorithm that can control the false negatives to a certain threshold is proposed. The theoretical foundation for this algorithm is based on Neyman-Pearson (NP) Lemma, which is used to construct a novel tree-based classifier that enables asymmetric error control. This classifier is evaluated on the data obtained from the Framingham heart study and it predicts the risk of a ten-year cardiac disease, not only with improved accuracy and F1 score but also with full control over the number of false negatives. With an improved accuracy in predicting cardiac disease, this tree-based classifier with asymmetric error control can reduce the burden of cardiac disease in populations and potentially save a lot of human lives. The methodology used to construct this classifier can be expanded to many more use cases in medical disease diagnosis.