Is this confusion matrix correct?
I'm trying to classify animal sounds. I'm using CNN for this purpose. I'm comparing the classes with each other, hence, some training will be done with 2 classes. Once I get the confusion matrix with these training, I encounter something like this:
As can be seen, it seems the model predicts everything correctly. On the other hand, when I raise the threshold from 0.5 to 0.8, I see that there are some misclassification. This is actually what I expected, but I confused that when I get the accuracy, precision, recall and f1 score, I see a value with lower than 1, and all are the same, as in the picture. However, I expect all of them as 1 because there is no misclassification. Am I wrong? Can you please explain? Thx in advance.
Note that my data has 2 class and each of which has 1200 records. I trained them with a CNN model. 20% of the data is used as test, rest of them are training.
Code is below:
from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score
from sklearn.metrics import confusion_matrix as cm
predictions = model.predict(X_test)
y_pred = (predictions > 0.5)
print(predictions.shape, y_pred.shape)
predictions = np.argmax(predictions, axis=1)
y_pred = np.argmax(y_pred, axis=-1)
cm(y_pred, y_test)
The problem is that you computed confusion matrix between prediction and itself, instead of prediction and y_test.
It should be
cm(y_pred, y_test)