评价分类结果（2）--精准度-召回率曲线、ROC曲线

1.精准度和召回率受决策边界的影响

导入数据

import numpy as np
import matplotlib.pyplot as plt

from sklearn import datasets

digits = datasets.load_digits()
X = digits.data
y = digits.target.copy()

y[digits.target==9] = 1
y[digits.target!=9] = 0

数据分割

from sklearn.model_selection import train_test_split

X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=666)

逻辑回归

from sklearn.linear_model import LogisticRegression

log_reg = LogisticRegression()
log_reg.fit(X_train, y_train)
y_predict = log_reg.predict(X_test)

各种衡量指标

# F1 Score
from sklearn.metrics import f1_score

f1_score(y_test, y_predict)
# 混淆矩阵
from sklearn.metrics import confusion_matrix

confusion_matrix(y_test, y_predict)
#精准率
from sklearn.metrics import precision_score

precision_score(y_test, y_predict)
#召回率
from sklearn.metrics import recall_score

recall_score(y_test, y_predict)

2.改变决策边界

逻辑回归中：以Xtheta == 0 为决策边界 Xtheta > 0 : 预测值为 1； X*theta < 0 : 预测值为 0

# 默认决策值边界为 0 
log_reg.decision_function(X_test)[:10] # 结果都小于0 

log_reg.predict(X_test)[:10] # 预测值都为 0 

# 保存决策值
decision_scores = log_reg.decision_function(X_test)

# 设置新的 决策边界 5 
y_predict_2 = np.array(decision_scores >= 5, dtype='int')

# 混合矩阵
confusion_matrix(y_test, y_predict_2)
# 精准率
precision_score(y_test, y_predict_2)
# 召回率
recall_score(y_test, y_predict_2)

# 设置新的 决策边界 -5
y_predict_3 = np.array(decision_scores >= -5, dtype='int')

# 混合矩阵
confusion_matrix(y_test, y_predict_3)
# 精准率
precision_score(y_test, y_predict_3)
# 召回率
recall_score(y_test, y_predict_3)

3.精准度-召回率曲线

from sklearn.metrics import precision_recall_curve

precisions,recalls,thresholds = precision_recall_curve(y_test,decision_scores)

import matplotlib.pyplot as plt
 
 
plt.rcParams['font.sans-serif']=['SimHei'] #用来正常显示中文标签
plt.rcParams['axes.unicode_minus']=False #用来正常显示负号

plt.plot(thresholds, precisions[:-1],label = "精准率")
plt.plot(thresholds, recalls[:-1],label = "召回率")
plt.legend()

plt.xlabel("决策边界值-thresholds", fontsize='15')
plt.ylabel("百分率", fontsize='15')

plt.show()

plt.plot(precisions, recalls)

plt.xlabel("precisions", fontsize='15')
plt.ylabel("recalls", fontsize='15')

plt.show()

4.ROC曲线

from sklearn.metrics import roc_curve

fprs, tprs, thresholds = roc_curve(y_test, decision_scores)
plt.plot(fprs, tprs)
plt.show()

ROC曲线包裹的面积：范围从[0,1] 越大越好.

from sklearn.metrics import roc_auc_score

roc_auc_score(y_test,decision_scores)