ROC Method for Identifying the Optimal Threshold With an Application to Email Classification
Oluwafunmibi Fasanya, Adetola A. Adediran, Olusegun Sunday Ewemooje, Femi B. Adebola · 2022
A cheap, fast, and reliable means of communication is email which is a widely used application by most people in their daily lives. For its easy navigation, several organizations and individuals choose to exploit this means of communication for sending out millions of advertisements at little or no cost. This, therefore, results in email users' inboxes being filled with all sorts of unsolicited and junk mails also called spam. To manage and classify these unsolicited emails, different machine learning algorithms with varying degree of success has been applied. Many of these algorithms were evaluated based on the preset threshold of 0.5. Thus, from the experiments conducted, this work showed that the Receiver Operating Curve–Area Under the Curve (ROC-AUC) is better suited to be used in the machine learning world when compared to the accuracy metrics which are generally used in assessing performance measure of a classification algorithm.