|
|
|
@@ -1,87 +0,0 @@ |
|
|
|
# -*- coding: utf-8 -*- |
|
|
|
# --- |
|
|
|
# jupyter: |
|
|
|
# jupytext_format_version: '1.2' |
|
|
|
# kernelspec: |
|
|
|
# display_name: Python 3 |
|
|
|
# language: python |
|
|
|
# name: python3 |
|
|
|
# language_info: |
|
|
|
# codemirror_mode: |
|
|
|
# name: ipython |
|
|
|
# version: 3 |
|
|
|
# file_extension: .py |
|
|
|
# mimetype: text/x-python |
|
|
|
# name: python |
|
|
|
# nbconvert_exporter: python |
|
|
|
# pygments_lexer: ipython3 |
|
|
|
# version: 3.5.2 |
|
|
|
# --- |
|
|
|
|
|
|
|
# # 混淆矩阵(confusion matrix) |
|
|
|
# |
|
|
|
# 混淆矩阵是用来总结一个分类器结果的矩阵。对于k元分类，其实它就是一个$k \times k$的表格，用来记录分类器的预测结果。 |
|
|
|
# |
|
|
|
# 对于最常见的二元分类来说，它的混淆矩阵是2乘2的，如下 |
|
|
|
#  |
|
|
|
# |
|
|
|
# * `TP` = True Postive = 真阳性 |
|
|
|
# * `FP` = False Positive = 假阳性 |
|
|
|
# * `FN` = False Negative = 假阴性 |
|
|
|
# * `TN` = True Negative = 真阴性 |
|
|
|
# |
|
|
|
# 你要的例子来了。。。比如我们一个模型对15个样本进行预测，然后结果如下。 |
|
|
|
# |
|
|
|
# * 预测值：1 1 1 1 1 0 0 0 0 0 1 1 1 0 1 |
|
|
|
# * 真实值：0 1 1 0 1 1 0 0 1 0 1 0 1 0 0 |
|
|
|
# |
|
|
|
#  |
|
|
|
# |
|
|
|
# |
|
|
|
# 这个就是混淆矩阵。混淆矩阵中的这四个数值，经常被用来定义其他一些度量。 |
|
|
|
# |
|
|
|
# |
|
|
|
# ### 准确度 |
|
|
|
# ``` |
|
|
|
# Accuracy = (TP+TN) / (TP+TN+FN+TN) |
|
|
|
# ``` |
|
|
|
# |
|
|
|
# 在上面的例子中，准确度 = (5+4) / 15 = 0.6 |
|
|
|
# |
|
|
|
# |
|
|
|
# |
|
|
|
# ### 精度(precision, 或者PPV, positive predictive value) |
|
|
|
# ``` |
|
|
|
# precision = TP / (TP + FP) |
|
|
|
# ``` |
|
|
|
# 在上面的例子中，精度 = 5 / (5+4) = 0.556 |
|
|
|
# |
|
|
|
# |
|
|
|
# |
|
|
|
# ### 召回(recall, 或者敏感度，sensitivity，真阳性率，TPR，True Positive Rate) |
|
|
|
# |
|
|
|
# ``` |
|
|
|
# recall = TP / (TP + FN) |
|
|
|
# ``` |
|
|
|
# |
|
|
|
# 在上面的例子中，召回 = 5 / (5+2) = 0.714 |
|
|
|
# |
|
|
|
# |
|
|
|
# |
|
|
|
# ### 特异度(specificity，或者真阴性率，TNR，True Negative Rate) |
|
|
|
# ``` |
|
|
|
# specificity = TN / (TN + FP) |
|
|
|
# ``` |
|
|
|
# |
|
|
|
# 在上面的例子中，特异度 = 4 / (4+2) = 0.667 |
|
|
|
# |
|
|
|
# |
|
|
|
# |
|
|
|
# ### F1-值(F1-score) |
|
|
|
# ``` |
|
|
|
# F1 = 2*TP / (2*TP+FP+FN) |
|
|
|
# ``` |
|
|
|
# 在上面的例子中，F1-值 = 2*5 / (2*5+4+2) = 0.625 |
|
|
|
# |
|
|
|
# |
|
|
|
# |
Shuhui Bu