在计算转移矩阵的时候，预测类别可能是实际类别的子集，sklearn自带的混淆矩阵不能直接使用，此处自定义，并进一步理解sklearn中混淆矩阵的定义。
混淆矩阵Matlab中X-Target, Y-Output (link)
Sklearn中X-Output, Y-Target (link)
- 代码
import numpy as np
from sklearn.metrics import confusion_matrix
def my_confusion_matrix(y_true, y_pred, cat_num = 2, normalize=None):
cm = np.zeros((cat_num, cat_num))
for (t,p) in zip(y_true, y_pred):
cm[t, p] += 1
if normalize is None:
pass
if normalize == 'pred':
cm = cm/cm.sum(axis=0, keepdims=True)
if normalize == 'true':
cm = cm/cm.sum(axis=1, keepdims=True)
if normalize == 'all':
cm = cm/np.sum(cm)
return cm
# Test 1
y_true = [0, 0, 0, 1, 1, 1, 1, 1]
y_pred = [0, 1, 0, 1, 0, 1, 0, 1]
cm_none = confusion_matrix(y_true, y_pred, normalize=None)
cm_pred = confusion_matrix(y_true, y_pred, normalize='pred')
cm_true = confusion_matrix(y_true, y_pred, normalize='true')
cm_all = confusion_matrix(y_true, y_pred, normalize='all')
print('None n', cm_none,'n')
print('Pred - Divide by sum of Col n', cm_pred,'n')
print('True - Divide by sum of Row n', cm_true,'n')
print('All- - Divide by sum of Mat n', cm_all, 'n')
# Test 2
cm_none = my_confusion_matrix(y_true, y_pred, normalize=None)
cm_pred = my_confusion_matrix(y_true, y_pred, normalize='pred')
cm_true = my_confusion_matrix(y_true, y_pred, normalize='true')
cm_all = my_confusion_matrix(y_true, y_pred, normalize='all')
print('None n', cm_none,'n')
print('Pred - Divide by sum of Col n', cm_pred,'n')
print('True - Divide by sum of Row n', cm_true,'n')
print('All- - Divide by sum of Mat n', cm_all, 'n')
- 结果
None [[2 1] [2 3]] Pred - Divide by sum of Col [[0.5 0.25] [0.5 0.75]] True - Divide by sum of Row [[0.66666667 0.33333333] [0.4 0.6 ]] All- - Divide by sum of Mat [[0.25 0.125] [0.25 0.375]] None [[2. 1.] [2. 3.]] Pred - Divide by sum of Col [[0.5 0.25] [0.5 0.75]] True - Divide by sum of Row [[0.66666667 0.33333333] [0.4 0.6 ]] All- - Divide by sum of Mat [[0.25 0.125] [0.25 0.375]]
Ref：
- https://scikit-learn.org/stable/modules/model_evaluation.html#confusion-matrix
