Throws error "IndexError: too many indices for array" when trying to plot roc for binary classification about scikit-plot HOT 4 CLOSED

@lugq1990 is correct. I never liked scikit-learn's roc_curve because I had to pick a positive class. I designed plot_roc_curve to work directly with the output of predict_proba.

Although I do realize now it could be useful for binary classifiers with decision_function, whose output is a 1D array. Perhaps somebody could extend the code of plot_roc_curve to work with a 1D array.

from scikit-plot.

In the scikit-plot, if you want to plot the ROC curve for a binary classification problem, the needed prob is a 2-dimentional array, maybe you just past a 1-D prob to the plot_roc_curve(). So you need to make the 1-D to 2-D just like this : np.concatenate((1-y_probas,y_probas),axis=1). Try this .

from scikit-plot.

@lugq1990 - ya, changing the numpy array to 2D worked. Thanks for your suggestion!

from scikit-plot.

This exact problem is occurring for me. I'll try @lugq1990 's solution. Thank you!

from scikit-plot.