Evaluation Metrics in Machine Learning - Confusion Matrix

Ji Kim·2021년 1월 13일

Machine Learning

목록 보기

Confusion Matrix which is often used as an evaluation metrics in binary classification shows how much the model is confused while performing the prediction. In other words, confusion matrix describes the performance of classification model on a set of test data for which the true values are known.

  • TN : Predicted Negative(0), Actually Negative(0)
  • FP : Predicted Positive(1), Actually Negative(0)
  • FN : Predicted Negative(0), Actually Positive(1)
  • TP : Predicted Positive(1), Actually Positive(1)

Scikit-learn provides confusion_matrix() API to visualize confusion matrix. Using the MyFakeClassifier from the previous post, we will pass pred and y_test as parameters of confusion_matrix().


# confusion matrix 
from sklearn.metrics import confusion_matrix

confusion_matrix(y_test, pred)


array([[405,   0],
       [ 45,   0]])

The confusion matrix does not have any TP or FP-values since the model simply predicted every values to be False(0).

Confusion matrix also provides multiple base-information to deduce various evaluation metrics such as accuracy, precision, and recall.

As we have studied previosuly, accuracy is determined by how much the model has correctly classified the feature data (TP + TN). Therefore, using the confusion matrix we can re-define the equation of accuracy as

Accuracy = (TP + TN) / (TP + TN + FP + FN)

if this then that

0개의 댓글