Which of the following metrics should a Machine Learning Specialist generally use to compare/evaluate machine learning classification models against each other?
RECALL IS ONE OF FACTOR IN CLASSIFY,
AUC IS MORE FACTORS TO COMPREHENSIVE JUDGEMENT
https://docs.aws.amazon.com/zh_tw/machine-learning/latest/dg/cross-validation.html
ANSWER MIGHT BE D.
Area Under the ROC Curve (AUC) is a commonly used metric to compare and evaluate machine learning classification models against each other. The AUC measures the model's ability to distinguish between positive and negative classes, and its performance across different classification thresholds. The AUC ranges from 0 to 1, with a score of 1 representing a perfect classifier and a score of 0.5 representing a classifier that is no better than random.
While recall is an important evaluation metric for classification models, it alone is not sufficient to compare and evaluate different models against each other. Recall measures the proportion of actual positive cases that are correctly identified as positive, but does not take into account the false positive rate.
D. AUC is always used to compare ML classification models. The others can all be misleading. Consider the cases where classes are highly imbalanced. In those cases accuracy, misclassification rate and the like are useless. Recall is only useful if used in combination with precision or specificity, which what AUC does.
AUC is to compare different models in terms of their separation power. 0.5 is useless as it's the diagonal line. 1 is perfect. I would go with F1 Score if it was an option. However, taking Recall only as a metric for comparing between models, would be misleading.
D. AUC is scale- and threshold-invariant, enabling it compare models.
https://towardsdatascience.com/how-to-evaluate-a-classification-machine-learning-model-d81901d491b1
Actually AUC could be generalized for multi-class problem.
https://www.datascienceblog.net/post/machine-learning/performance-measures-multi-class-problems/
Could be, you mean in a multiclass clasification problem. But in that con context recall directly can't be compare because first you have to decide recall of what of the classes, in a 3 classes problem we have 3 recalls or you suppose a weighted recall or average recall ?. Do you think in that ?
Also in multi-class classification, if you follow an One-vs_Rest strategy you can still use AUC.
https://scikit-learn.org/stable/auto_examples/model_selection/plot_roc.html#sphx-glr-auto-examples-model-selection-plot-roc-py
Correct Answer is D. Another benefit of using AUC is that it is classification-threshold-invariant like log loss.
https://towardsdatascience.com/the-5-classification-evaluation-metrics-you-must-know-aa97784ff226
A voting comment increases the vote count for the chosen answer by one.
Upvoting a comment with a selected answer will also increase the vote count towards that answer by one.
So if you see a comment that you already agree with, you can upvote it instead of posting a new comment.
DonaldCMLIN
Highly Voted 3 years, 3 months agodevsean
3 years, 3 months agoDScode
3 years, 2 months agoAjoseO
Highly Voted 1 year, 10 months agoccpmad
1 year, 5 months agoAsusTuf
Most Recent 1 year, 2 months agoScrook
7 months, 2 weeks agoMickey321
1 year, 4 months agoValcilio
1 year, 9 months agocloud_trail
3 years, 1 month agoharmanbirstudy
3 years, 2 months agoMohamedSharaf
3 years, 1 month agoharmanbirstudy
3 years, 2 months agoDavidRou
1 year, 3 months agoThai_Xuan
3 years, 2 months agojohnny_chick
3 years, 2 months agodeep_n
3 years, 2 months agohughhughhugh
3 years, 2 months agoPRC
3 years, 2 months agoHypermasterd
3 years, 2 months agooMARKOo
3 years, 2 months agosebas10
3 years, 2 months agomrsimoes
3 years, 2 months agostamarpadar
3 years, 3 months ago