Skip to content

Confused?

1904 (Origin): Karl Pearson

As artificial intelligence grew, the "confusion matrix" became a standard tool for evaluating classification models.

What we get from confusion metric?

Note

TP - Model sahi tha aur model ne response me P (Positive kaha tha) aur actually me Positive tha
TN - Model sahi tha aur model ne response me N (Negative kaha tha) aur actually me Negative tha
FP - Model galat tha aur model ne response me P (Positive kaha tha) lekin actually me Negative tha
FN - Model galat tha aur model ne response me N (Negative kaha tha) lekin actually me Positive tha

Below are the possibilities of representing a confusion matrix

Predicterd on X-axis with first col as NO
And actual as Y-axis with first col as NO

Predicted: NO/0 Predicted: YES/1
Actual: NO/0 True Negative (TN) False Positive (FP)
Actual: YES/1 False Negative (FN) True Positive (TP)

Predicterd on X-axis with first col as YES
And actual as Y-axis with first col as YES

Predicted: YES/1 Predicted: NO/0
Actual: YES/1 True Positive (TP) False Negative (FN)
Actual: NO/0 False Positive (FP) True Negative (TN)

Predicterd on Y-axis with first col as NO
And actual as X-axis with first col as NO

Actual: NO/0 Actual: YES/1
Predicted: NO/0 True Negative (TN) False Negative (FN)
Predicted: YES/1 False Positive (FP) True Positive (TP)

Predicterd on Y-axis with first col as YES
And actual as X-axis with first col as YES

Actual: YES/1 Actual: NO/0
Predicted: YES/1 True Positive (TP) False Positive (FP)
Predicted: NO/0 False Negative (FN) True Negative (TN)

How to understand all these 4 - TP, TN, FP, FN

Dekho First letter - T or F

T or F

First letter ye represent karta hai ki model ka response sahi hua ki galat

Ques: Sahi kab hoga?
Ans: Jab model ne jo kaha wo actual me jo hai usase match karta ho

Sahi - T
Galat - F

Second letter - P or N

P or N

Second letter ye represent karta hai ki model ne apane response me kya kaha (Positive(➕) kaha ya Negative(➖))

Positive - P
Nevative - N

Summary

TP - If this is higher - model is doing good at predicting Positive cases
TN - If this is higher - model is doing good at predicting Negative cases
FP - Many false positives → model is crying “YES!” too often
FN - Many false negatives → model is missing real cases

From this confusion metric we get over 10 useful key metric

  1. Accuracy
  2. Precision
  3. Recall
  4. Specificity
  5. F1-Score