Azure Series: AI — Basics of Machine Learning, Defining True Positive (TP), False Positive (FP), False Negative ( FN), and True Negative (TN) (Introduction)

This article introduces important concepts that are used to measure classification models.

Example Wolf Identification Solution

A farmer experience a problem of a wolf eating her sheep. To address the problem the farmer considers installing cameras that will use machine learning to identify the wolf. The farmer has a dog, the machine learning must be able to differentiate between wolf and dog. When the machine learning identifies the wolf it should trigger the notification that is sent to the farmer's smartphone.

Defining TP, FP, FN, and TN

The example above highlights the wolf, dog, and notification. We are going to use those words to easily define TP, FP, FN, and TN.

“Wolf” is a Positive class

“Dog” is a Negative class

“Notification” is a trigger that is sent when a wolf is identified

WIS is Wolf Identification Solution

True Positive: an outcome where WIS accurately identifies a wolf and sends a notification to the farmer

False Positive: An outcome where the WIS machine learning model incorrectly identifies a dog as the wolf and sends a notification to the farmer that it saw a wolf

False Negative: An outcome where the WIS machine learning model incorrectly identifies a wolf as a dog and doesn’t send the notification.

True Negative: An outcome where the WIS machine learning model correctly identifies a dog and doesn’t send the notification.

Summary

In short, this article explained the common concept that is measuring the performance of machine learning models.