Precision, Recall, & F1 Score Intuitively Explained

preview_player
Показать описание
Classification performance metrics are an important part of any machine learning system. Here we discuss the most basic and common measures of model performance; accuracy, precision, recall, and F1 score.

Chapters:
0:00 Introduction
1:02 Basic Definitions
2:50 Accuracy
4:30 Precision
5:36 Recall
6:34 F1 Score
8:07 Conclusion
  1. Scarlett's Log
Рекомендации по теме
Never Forget Again! 0:05:24

Never Forget Again! // Precision vs Recall with a Clear Example of Precision and Recall

Precision, Recall, F1 0:11:46

Precision, Recall, F1 score, True Positive|Deep Learning Tutorial 19 (Tensorflow2.0, Keras & Pyt...

Confusion Matrix Solved 0:05:50

Confusion Matrix Solved Example Accuracy Precision Recall F1 Score Prevalence by Mahesh Huddar

Introduction to Precision, 0:03:06

Introduction to Precision, Recall and F1 | Classification Models

Precision, Recall, & 0:08:56

Precision, Recall, & F1 Score Intuitively Explained

Precision, Recall and 0:09:03

Precision, Recall and F1-Score

Como entender as 0:11:36

Como entender as métricas Precision & Recall

8.8. Precision, Recall, 0:32:12

8.8. Precision, Recall, F1 score | Model Evaluation

Machine Learning Fundamentals: 0:07:13

Machine Learning Fundamentals: The Confusion Matrix

How to evaluate 0:10:05

How to evaluate ML models | Evaluation metrics for machine learning

TP, FP, TN, 0:14:01

TP, FP, TN, FN, Accuracy, Precision, Recall, F1-Score, Sensitivity, Specificity, ROC, AUC

MFML 044 - 0:05:47

MFML 044 - Precision vs recall

Precision, Recall, and 0:09:04

Precision, Recall, and F1 Score Explained for Binary Classification

Precision e recall 0:07:29

Precision e recall

Precision und Recall 0:01:41

Precision und Recall in 100 Sekunden

Precision, Recall and 0:42:42

Precision, Recall and F1 Score | Classification Metrics Part 2

Scikit-Learn Classification Report 0:00:32

Scikit-Learn Classification Report - Precision, Recall, F1, Accuracy of ML Models

What are Precision 0:08:08

What are Precision and Recall in Machine Learning?

Precision and Recall 0:14:30

Precision and Recall | Unforgettable explanation with easy example on White Board | Confusion Matrix

Confusion Matrix for 0:08:22

Confusion Matrix for Multiclass Classification Precision Recall Weighted F1 Score by Mahesh Huddar

12.2 Precision, Recall, 0:11:48

12.2 Precision, Recall, and F1 Score (L12 Model Eval 5: Performance Metrics)

Precision, Recall, F1-score, 0:12:43

Precision, Recall, F1-score, confusion matrix, multilabel classification micro e macro average (ita)

Mastering the Confusion 0:06:08

Mastering the Confusion Matrix | Accuracy, Precision, Recall & F1 Score | Machine Learning

Tutorial 34- Performance 0:24:12

Tutorial 34- Performance Metrics For Classification Problem In Machine Learning- Part1

Комментарии
Автор

Many thanks for this intuitive yet detailed and well-reasoned explanation!

swedishguyonyoutube
Автор

Awesome explanation, thank you very much. It shines because it gives the motivation for each metric. Thanks!

jhfoleiss
Автор

These concepts "clicked" for me thanks to this video, really great work and thank you so much for sharing.

Maddy
Автор

Another good explanation of F1 score is that it considers FN and FP equally like:
Tp/(TP + 1/2(FN+FP)). So this is another way how it's a goos mix of precision and recall.

sarvagyagupta
Автор

Probably the best explanation on the net.

justsuyash
Автор

Great explanation because it includes motivation for the metrics and how they relate to each other.

vladimirblagojevic
Автор

Great explanation, thank you very much

RolandoLopezNieto
Автор

I couldnt get one thing, putting here if you are able to answer please ---> ' in the Slide of 'Precision', you mentioned that it solves problem with Accuracy cheating, where if 99% of people are healthy and 1% unhealthy, and if your model predict all as 'Healthy' then Precision would be 0, but how ? lets say 100 people, 99 healthy(positive), 1 unhealthy (negative), prediction made by model as All Positive (cheated), thus 99 (True-Positive), 1 ( False-Positive), Precision = TP / ( TP +FP ) = 99 / ( 99 + 1 ) = 0.99 = 99% right ? how 0 ?

samirkhan
Автор

Nice explanations. When you say negative labels are left out of picture in the precision calculation, I don’t quite get it because precision formula contains the term FP which represents a quantity from negative labels. Am I out of the track?

janithforex
Автор

ENJOYED THE VIDEO - keep up the great work !
I DO find myself with one question though... what about the "NULL" case did it actually get considered?

In one example you describe a type B as the "positive" case amongst types A, B, C and D; this makes SENSE, BUT...
... you then describe the "negatives" as being ONLY A, C and D... but what about the case of "NOT A, B, C OR D" ?
This is the "null" set, and it occurs quite regularly in Computer Vision OBJECT DETECTION models.

IMAGINE THIS SCENARIO:
An image of FRUIT on a TABLE - 3 red apples, 1 yellow apple, and 2 bananas.

SAY you're looking to detect APPLES, but your MODEL only detects the 3 RED APPLES and UTTERLY MISSES the 1 YELLOW APPLE?
THIS is a case of there BEING an APPLE, but you DIDN'T MIS-CLASSIFY it as the BANANA... YOU MISSED DETECTING IT COMPLETELY !

HOW would you describe THIS common scenario if you're only considering/assuming that your model WILL classify EVERYTHING in the image as EITHER an APPLE or a BANANA... but you DIDN'T expect it to UTTERLY IGNORE the yellow apple altogether?

It's been 2 years since you posted, so I'm not expecting a reply; I AM hoping that other viewers will ponder the explanations presented - there's a bit more going on...

Cheers,
-Mark Vogt, Solution Architect/Data Scientist - AVANADE

VOGTLANDOUTDOORS
Автор

Awesome! I love the cheating cases brought in!

elvenkim