CODEWITHSUNDEEP
pythonscikit-learnconfusion-matrix
Economister
Economister

Reputation: 67

Why does my sklearn.metrics confusion_matrix output look transposed?

It's my understanding that confusion matrices should show the TRUE classes in the columns and the PREDICTED classes in the rows. Therefore the sum of the columns should be equal to the value_counts() of the TRUE series.

I have provided an example here:

from sklearn.metrics import confusion_matrix

pred = [0, 0, 0, 1]
true = [1, 1, 1, 1]

confusion_matrix(true, pred)

Why does this give me the following output? Surely it should be the transpose of that?

array([[0, 0],
       [3, 1]], dtype=int64)

Upvotes: 1

Views: 2347

Answers (3)

Ivanovitch Silva
Ivanovitch Silva

Reputation: 1

It is possible to do as you wish using sklearn, only change the code below appropriately

from sklearn.metrics import ConfusionMatrixDisplay
import matplotlib.pyplot as plt

fig, ax = plt.subplots(1,1,figsize=(7,4))

ConfusionMatrixDisplay(confusion_matrix(predict,y_test,labels=[1,0]),
                       display_labels=[1,0]).plot(values_format=".0f",ax=ax)

ax.set_xlabel("True Label")
ax.set_ylabel("Predicted Label")
plt.show()

Upvotes: 0

Sunil
Sunil

Reputation: 11

The confusion probably arises because sklearn follows a different convention for axes of confusion matrix than the wikipedia article. So, to answer your question: It gives you the output in that specific format because sklearn expects you to read it in a specific way.

Here are the two different ways of writing confusion matrix:

Upvotes: 1

seralouk
seralouk

Reputation: 33147

scikit-learn's confusion matrix follows a specific order and structure.

Reference: https://scikit-learn.org/stable/auto_examples/model_selection/plot_confusion_matrix.html#sphx-glr-auto-examples-model-selection-plot-confusion-matrix-py

enter image description here

Upvotes: 0

Related Questions