Support Vector Machines and ROCAUC

Describe the bug ROCAUC fails with support vector machines. This doesn’t happen to me with other algorithms.

To Reproduce

model = SVC()
viz = ROCAUC(model, n_jobs=-1)
viz.fit(X_train, y_train)
viz.score(X_test, y_test)

Dataset I’m using the credit dataset on Yellowbrick

Expected behavior To obtain the ROC plot

/media/Data/Gatech/machine-learning/supervised-learning/helpers.py in plt_roc(model, X_train, y_train, X_test, y_test)
     30     viz = ROCAUC(model, n_jobs=-1)
     31     viz.fit(X_train, y_train)
---> 32     viz.score(X_test, y_test)
     33     return viz
     34 

~/venvs/global/lib/python3.6/site-packages/yellowbrick/classifier/rocauc.py in score(self, X, y, **kwargs)
    176         # Compute ROC curve and ROC area for each class
    177         for i, c in enumerate(classes):
--> 178             self.fpr[i], self.tpr[i], _ = roc_curve(y, y_pred[:,i], pos_label=c)
    179             self.roc_auc[i] = auc(self.fpr[i], self.tpr[i])

IndexError: too many indices for array

Desktop (please complete the following information):

• OS: Ubuntu 18.04
• Python Version: 3.6.5
• Yellowbrick Version: 0.8

[FranGoitia]

Thank you for fixing it and working on such a fantastic library ! Best, Fran

[bbengfort]

Thanks @TaiChiTiger for your note!