ABSTRACT
Receiver operating characteristic (ROC) curve analysis is a popular tool for visualizing, organizing, and selecting classifiers based on their classification accuracy. The ROC curve methodology was originally developed during World War II to analyze classification accuracy in differentiating signal from noise in radar detection. ROC curves are increasingly used in the machine learning field, due in part to the realization that simple classification accuracy is often a poor standard for measuring performance. In addition to being a generally useful graphical method to visualize classification accuracy, ROC curves have properties that make them especially useful for domains with skewed discriminating distributions or unequal classification error costs. ROC analysis can also be applied generally for evaluating the accuracy of goodness of fit of a statistical model (e.g., logistic regression and linear discriminant analysis) that classifies subjects into two categories, such as diseased or nondiseased.
