TypeError: plot_confusion_matrix() got an unexpected keyword argument 'ax'

[Schmidtbit]

plot_confusion_matrix() doesn't appear to accept an axis argument. However the documentation does indicate that you can pass in an axis.

I am trying to plot on an existing figure into subplots, but getting this error:

TypeError: plot_confusion_matrix() got an unexpected keyword argument 'ax'

[rasbt]

could you try plot_confusion_matrix(... , axis=...) instead of plot_confusion_matrix(... , ax=...) and see if that works?

[Schmidtbit]

I did as you said and got this new error:

/opt/anaconda3/envs/py37/lib/python3.7/site-packages/mlxtend/plotting/plot_confusion_matrix.py in plot_confusion_matrix(conf_mat, hide_spines, hide_ticks, figsize, cmap, colorbar, show_absolute, show_normed, class_names, figure, axis)
     84         fig, ax = figure, axis
     85 
---> 86     ax.grid(False)
     87     if cmap is None:
     88         cmap = plt.cm.Blues

AttributeError: 'numpy.ndarray' object has no attribute 'grid'```

[rasbt]

Based on the error, it sounds like you provided a numpy array instead of axis object. The following code example would work:

from mlxtend.plotting import plot_confusion_matrix
import matplotlib.pyplot as plt
import numpy as np

binary1 = np.array([[4, 1],
                   [1, 2]])

fig, axis = plt.subplots(1, 1)

fig, ax = plot_confusion_matrix(conf_mat=binary1, axis=axis)
plt.show()

if you have something like

fig, axis = plt.subplots(1, 2)

then "axis" will be a numpy array of multiple axis. But since the confusion matrix is just a single plot, it would be unclear which axis to use. You could select a specific one, e.g, by axis=axis[0] in that case or axis=axis[1]

[Schmidtbit]

That worked! Thanks!

But in order for it to work I had to turn off the colorbar. Otherwise I got:

/opt/anaconda3/envs/py37/lib/python3.7/site-packages/mlxtend/plotting/plot_confusion_matrix.py in plot_confusion_matrix(conf_mat, hide_spines, hide_ticks, figsize, cmap, colorbar, show_absolute, show_normed, class_names, figure, axis)
     97 
     98     if colorbar:
---> 99         fig.colorbar(matshow)
    100 
    101     for i in range(conf_mat.shape[0]):

AttributeError: 'NoneType' object has no attribute 'colorbar'

[Schmidtbit]

However, the axis labels only show up for the last subplot.

fig, ax = plt.subplots(1,3,figsize=(20,10))
for i, clf in enumerate(['DeepNeuralNetwork','RandomForest','GradientBoosted']):
    plot_confusion_matrix(conf_mat=cm_arr[clf],
                        colorbar=False,
                        show_absolute=False,
                        show_normed=True,
                        class_names=traindf['PartStatus'].unique(),
                        axis=ax[i])
    ax[i].set_title(clf)
plt.show()

[Schmidtbit]

I was able to get around the formatting with:

fig, ax = plt.subplots(1,3,figsize=(25,10),sharex=True, sharey=True)
fig.text(0.5, 0.04, 'True Label', ha='center',fontsize=20)
fig.text(0.04, 0.5, 'Predicted Label', va='center', rotation='vertical',fontsize=20)
for i, clf in enumerate(['DeepNeuralNetwork','RandomForest','GradientBoosted']):
    plot_confusion_matrix(conf_mat=cm_arr[clf],
                        colorbar=False,
                        show_absolute=False,
                        show_normed=True,
                        class_names=None,
                        axis=ax[i])
    ax[i].set_title(clf, fontsize=20)
    ax[i].set_xticks([0,1,2])
    ax[i].set_xticklabels(traindf['PartStatus'].unique())
    ax[i].set_yticks([0,1,2])
    ax[i].set_yticklabels(traindf['PartStatus'].unique()[::-1])
    ax[i].set_ylabel('')
    ax[i].set_xlabel('')
plt.show()

[rasbt]

Glad to hear that you got it to work. Looks really nice. We should probably add something like this to the documentation as an example for how to make a plot with multiple subpanels.