Demystifying the Receiver Operating Characteristic (ROC) Curve

Author(s): Tan Pengshi Alvin

Originally published on Towards AI.

Interpreting the metrics behind a supervised classification model, in particular, the ROC curve and AUC
Image by Chris Martin from Pixabay

One of the most common and widely applicable themes in Artificial Intelligence is Supervised Learning, in particular the Classification task. For instance, if we have a mixture of cat and dog images in a dataset, the process of labeling them as ‘cats’ or ‘dogs’ (and converting them to the binary values 0 or 1) involves human supervision. Hence, the Supervised Learning classification model is an optimized function that maps all the pixel values (0 to 255 in each RGB channel) of each image to the supervised label 0 or 1. In essence, supervised classification is about predicting the category or class that each presented data point (in this case, an image) belongs to.

The process of model optimization and image classification is extensively discussed in the following articles, and I will leave the reader to further explore:

A complete guide from CNN to Transfer Learning, with application on Kaggle’s Cat versus Dog dataset

tanpengshi.medium.com

Developing deep neural networks from scratch with Mathematics and Python

tanpengshi.medium.com

This article instead focuses on understanding the metrics of model evaluation for Classification, in particular, it aims to offer a complete and intuitive interpretation of the Receiver Operating Characteristic (ROC) Curve and Area Under Curve (AUC),… Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI