Cannot get classification report even after rounding up predictions

Question

I am trying to classify my dataset using keras but I am getting ValueError: Classification metrics can't handle a mix of multiclass and multilabel-indicator targets error. values in y_pred are as following

array([[2.95522604e-02, 9.70325887e-01, 3.20542094e-05, ...,
        1.74383260e-07, 1.98587145e-07, 9.88743452e-08],
       [3.25102806e-01, 6.68996394e-01, 1.65001326e-03, ...,
        5.84201662e-05, 5.91963508e-05, 4.68929684e-05],
       [8.87618303e-01, 1.12024814e-01, 1.22764613e-04, ...,
        1.44616331e-06, 1.33618846e-06, 1.68983024e-06],
       ...,
       [3.09438616e-01, 6.83520675e-01, 1.94711238e-03, ...,
        7.57295784e-05, 7.51852640e-05, 5.94857411e-05],
       [6.73729360e-01, 3.21534157e-01, 1.41171378e-03, ...,
        4.93246625e-05, 4.61974196e-05, 4.73670734e-05],
       [1.33120596e-01, 8.64127636e-01, 7.41749362e-04, ...,
        1.87505502e-05, 1.95825924e-05, 1.34223355e-05]], dtype=float32)

I am rounding them up as mentioned in this question as y_test values are

array([1, 0, 0, ..., 0, 1, 1])

After rounding y_pred with y_pred = y_pred.round().astype(int) I have

array([[0, 1, 0, ..., 0, 0, 0],
       [1, 0, 0, ..., 0, 0, 0],
       [1, 0, 0, ..., 0, 0, 0],
       ...,
       [0, 1, 0, ..., 0, 0, 0],
       [1, 0, 0, ..., 0, 0, 0],
       [0, 1, 0, ..., 0, 0, 0]])

Bit even after this when i try to get classification report using print(metrics.classification_report(y_test, y_pred)) I get same error as above mentioned. Can someone help me about what am I doing wrong here? Thank you

score 2 · Accepted Answer · answered Feb 23 '21 at 21:53

The scikit-learn docs states that the y_pred input must be a 1d array-like. You need to argmax your logits.

import numpy as np
import tensorflow as tf
from sklearn.metrics import classification_report


y_pred = tf.math.abs(tf.random.normal([32, 2])).numpy()
y_test = tf.random.uniform([32, 1], minval=0, maxval=2, dtype=tf.int32).numpy()

# this will explode
print(classification_report(y_test, y_pred))

# ValueError: Classification metrics can't handle a mix of binary and 
# continuous-multioutput targets

# get predicted indices
y_pred = np.argmax(y_pred, 1)

# try again
print(classification_report(y_test, y_pred))

#                precision    recall  f1-score   support
#
#             0       0.41      0.50      0.45        14
#             1       0.53      0.44      0.48        18
# 
#      accuracy                           0.47        32
#     macro avg       0.47      0.47      0.47        32
#  weighted avg       0.48      0.47      0.47        32

Cannot get classification report even after rounding up predictions

1 Answers1

Linked