Abstract

In a classification problem using statistical inference, the predictions from a model are compared with the actual observed values (the "ground truth") and represented in an array called the confusion matrix. This matrix has been widely used to draw inferences in both the frequentist and Bayesian frameworks. In the frequentist approach, various metrics of the confusion matrix, viz. accuracy, precision, sensitivity, specificity, F1-score etc., can be directly computed from the observed data without the need to compute the prior