Plot a confusion matrix.

plotcm.py

import numpy as np
import matplotlib.pyplot as plt
import matplotlib as mpl
import matplotlib.patches as patches

mpl.rcParams['font.family'] = "Times New Roman"
#mpl.rcParams['font.size'] = "11"
mpl.rc('pdf', fonttype=42)


cm = [['en', '0.842', '0.508', '0.425', '0.471', '0.316', '0.252', '0.166', '0.291'], ['es', '0.628', '0.662', '0.349', '0.445', '0.29', '0.125', '0.019', '0.36'], ['de', '0.683', '0.491', '0.505', '0.455', '0.312', '0.353', '0.265', '0.289'], ['nl', '0.666', '0.528', '0.407', '0.563', '0.308', '0.151', '0.047', '0.299'], ['tr', '0.657', '0.518', '0.341', '0.411', '0.472', '0.231', '0.118', '0.358'], ['uz', '0.565', '0.446', '0.367', '0.391', '0.332', '0.481', '0.305', '0.265'], ['bn', '0.604', '0.442', '0.236', '0.316', '0.314', '0.418', '0.363', '0.309'], ['ha', '0.583', '0.358', '0.255', '0.29', '0.266', '0.123', '0.059', '0.403']]

langs = map(lambda l: l[0], cm)
cm = map(lambda l: map(lambda n: 100*float(n), l[1:]), cm)


def plot_confusion_matrix(cm, title='Confusion matrix'):
    plt.imshow(cm, interpolation='nearest', cmap=plt.get_cmap("YlGn"))
    #plt.colorbar()
    tick_marks = np.arange(len(langs))
    plt.xticks(tick_marks, langs, rotation='vertical')
    plt.yticks(tick_marks, langs)

    plt.tick_params(length=0)

    plt.ylabel('Train', fontsize=18)
    ax = plt.gca()
    ax.xaxis.tick_top()
    ax.set_xlabel('Test', fontsize=18)
    ax.xaxis.set_label_position('top')

    npcm = np.array(cm)

    bests = []
    nextbests = []
    for i,col in enumerate(npcm.T):
        best = np.argsort(col)[-1]
        nextbest = np.argsort(col)[-2]
        bests.append((i,best))
        nextbests.append((i,nextbest))


    for x in xrange(len(cm)):
        for y in xrange(len(cm)):
            plt.annotate(str(cm[x][y]), xy=(y, x),
                        horizontalalignment='center',
                        verticalalignment='center')

            # if max in column, also draw rect.

            # if (x,y) in bests:
            #     ax.add_patch(
            #             patches.Rectangle(
            #                     (x-0.5, y-0.5),
            #                     1,
            #                     1,
            #                     fill=False,
            #                     edgecolor="red",
            #                     linewidth=3
            #             )
            #     )

            if (x,y) in nextbests:
                ax.add_patch(
                        patches.Rectangle(
                                (x-0.5, y-0.5),
                                1,
                                1,
                                fill=False,
                                edgecolor="red",
                                linewidth=4
                        )
                )


# Compute confusion matrix
np.set_printoptions(precision=2)

plt.figure()
plot_confusion_matrix(cm, title='Confusion matrix')

plt.savefig("confmat.pdf", format='pdf')
plt.show()