Machine Learning - Confusion Matrix

1 - About

If a classification system has been trained a confusion matrix will summarize the results (ie the error rate (false|true) (positive|negative) for a binary classification).

This is training error, there is may be overfitting.

The main diagonal indicates correct classification whereas everything off the main diagonal indicates a classification.

Predicted class
Actual Class Yes No
Yes True Positive False Negative
No False Positive True Negative

3 - Example

3.1 - Titanic Data Set

With a simple rule: If most females survived, then assume every female survives

Survived: Yes or No

Predicted class
Actual Class Yes No
Yes 233 81
No 109 468
  • Total of good predictions : 233 + 468 = 701
  • Total of bad predictions : 81 + 109 = 190
  • Percentage of good prediction: 701 /( 701 + 190)*100 = 79%

4 - Documentation / Reference

data_mining/confusion_matrix.txt · Last modified: 2015/06/24 21:18 by gerardnico