Scoring categorical class data

WWD — Mon, 18 Oct 2021 15:53:17 GMT

I have a question about a topic within the following area:

Course: AI and Machine Learning Professional

Module: Natural Language and Computer Vision Professional

Lesson 5

Video: Scoring Categories Part 2

Time Position that produced the question 6:29

In this video, the instructor uses a suite of supervised learning techniques to predict a certain categorical variable. After all the calculations are made, the diagnostic summary statistics and graphs are presented. Within this material, the ROC curve is presented.

In the video, the ROC curve for this example is NOT non-decreasing. I always thought that the ROC curve always began at point (0,0) and ended at point (1,1). In the ROC curve for this example, the ROC curve doesn't extend out to 1,1 AND actually decreases as the x variable increases from a certain point.

Please help me understand the ROC curve in the demonstration.

Thank you,

Bill Donaldson

Re: Scoring categorical class data

Rick_SAS — Fri, 13 May 2022 18:38:56 GMT

Yes, you are correct about ROC curves (for binary classification) being nondecreasing. However, there are other curves that appear in machine learning. You might be seeing a lift curve or some other chart in which the curve can decrease.

topic Scoring categorical class data in SAS Academy for Data Science

Scoring categorical class data

Re: Scoring categorical class data