Solved – How to interpret a confusion matrix

confusion matrixpredictionpredictive-models

I am using confusion matrix to check the performance of my classifier.

I am using Scikit-Learn, I am little bit confused. How can I interpret the result from 

from sklearn.metrics import confusion_matrix
>>> y_true = [2, 0, 2, 2, 0, 1]
>>> y_pred = [0, 0, 2, 2, 0, 2]
>>> confusion_matrix(y_true, y_pred)
array([[2, 0, 0],
       [0, 0, 1],
       [1, 0, 2]])

How can I take the decision whether this predicted values are good or no.

Best Answer

The confusion matrix is a way of tabulating the number of misclassifications, i.e., the number of predicted classes which ended up in a wrong classification bin based on the true classes.

While sklearn.metrics.confusion_matrix provides a numeric matrix, I find it more useful to generate a 'report' using the following:

import pandas as pd
y_true = pd.Series([2, 0, 2, 2, 0, 1, 1, 2, 2, 0, 1, 2])
y_pred = pd.Series([0, 0, 2, 1, 0, 2, 1, 0, 2, 0, 2, 2])

pd.crosstab(y_true, y_pred, rownames=['True'], colnames=['Predicted'], margins=True)

which results in:

Predicted  0  1  2  All
True                   
0          3  0  0    3
1          0  1  2    3
2          2  1  3    6
All        5  2  5   12

This allows us to see that:

  1. The diagonal elements show the number of correct classifications for each class: 3, 1 and 3 for the classes 0, 1 and 2.
  2. The off-diagonal elements provides the misclassifications: for example, 2 of the class 2 were misclassified as 0, none of the class 0 were misclassified as 2, etc.
  3. The total number of classifications for each class in both y_true and y_pred, from the "All" subtotals

This method also works for text labels, and for a large number of samples in the dataset can be extended to provide percentage reports.

import numpy as np
import pandas as pd

# create some data
lookup = {0: 'biscuit', 1:'candy', 2:'chocolate', 3:'praline', 4:'cake', 5:'shortbread'}
y_true = pd.Series([lookup[_] for _ in np.random.random_integers(0, 5, size=100)])
y_pred = pd.Series([lookup[_] for _ in np.random.random_integers(0, 5, size=100)])

pd.crosstab(y_true, y_pred, rownames=['True'], colnames=['Predicted']).apply(lambda r: 100.0 * r/r.sum())

The output then is:

Predicted     biscuit  cake      candy  chocolate    praline  shortbread
True                                                                    
biscuit     23.529412    10  23.076923  13.333333  15.384615    9.090909
cake        17.647059    20   0.000000  26.666667  15.384615   18.181818
candy       11.764706    20  23.076923  13.333333  23.076923   31.818182
chocolate   11.764706     5  15.384615   6.666667  15.384615   13.636364
praline     17.647059    10  30.769231  20.000000   0.000000   13.636364
shortbread  17.647059    35   7.692308  20.000000  30.769231   13.636364

where the numbers now represent the percentage (rather than number of cases) of the outcomes that were classified.

Although note, that the sklearn.metrics.confusion_matrix output can be directly visualized using:

import matplotlib.pyplot as plt
conf = sklearn.metrics.confusion_matrix(y_true, y_pred)
plt.imshow(conf, cmap='binary', interpolation='None')
plt.show()
Related Question