How to fix ValueError: multiclass format is not supported

Question

This is my code and I try to calculate ROC score but I have a problem with ValueError: multiclass format is not supported. I'm already looking sci-kit learn but it doesn't help. In the end, I'm still have ValueError: multiclass format is not supported.

This is my code

from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import BaggingClassifier
from sklearn.metrics import confusion_matrix,zero_one_loss
from sklearn.metrics import classification_report,matthews_corrcoef,accuracy_score
from sklearn.metrics import roc_auc_score, auc


dtc = DecisionTreeClassifier()
bc = BaggingClassifier(base_estimator=dtc, n_estimators=10, random_state=17)
bc.fit(train_x, train_Y)
pred_y = bc.predict(test_x)

fprate, tprate, thresholds = roc_curve(test_Y, pred_y)
results = confusion_matrix(test_Y, pred_y)
error = zero_one_loss(test_Y, pred_y)
roc_auc_score(test_Y, pred_y)

FP = results.sum(axis=0) - np.diag(results)  
FN = results.sum(axis=1) - np.diag(results)
TP = np.diag(results)
TN = results.sum() - (FP + FN + TP)



print('\n Time Processing: \n',time.process_time())
print('\n Confussion Matrix: \n', results)
print('\n Zero-one classification loss: \n', error)
print('\n True Positive: \n', TP)
print('\n True Negative: \n', TN)
print('\n False Positive: \n', FP)
print('\n False Negative: \n', FN)
print ('\n The Classification report:\n',classification_report(test_Y,pred_y, digits=6))
print ('MCC:', matthews_corrcoef(test_Y,pred_y))
print ('Accuracy:', accuracy_score(test_Y,pred_y))
print (auc(fprate, tprate))
print ('ROC Score:', roc_auc_score(test_Y,pred_y))

This is the traceback

Solution

From the docs, roc_curve: "Note: this implementation is restricted to the binary classification task."

Are your label classes (y) either 1 or 0? If not, I think you have to add the pos_label parameter to your roc_curve call.

fprate, tprate, thresholds = roc_curve(test_Y, pred_y, pos_label='your_label')

Or:

test_Y = your_test_y_array  # these are either 1's or 0's
fprate, tprate, thresholds = roc_curve(test_Y, pred_y)