# Final-决策树（第二问-混淆矩阵、P、R和F1值，作出P-R曲线、ROC曲线，并求AUC）
from sklearn.model_selection import StratifiedKFold
import pandas as pd
from sklearn import tree
from sklearn.metrics import accuracy_score
import numpy as np
import matplotlib.pyplot as plt
from sklearn.metrics import precision_recall_curve
def PR_curve(y, pred):
pos = np.sum(y == 1)
neg = np.sum(y == 0)
pred_sort = np.sort(pred)[::-1] # 从大到小排序
index = np.argsort(pred)[::-1] # 从大到小排序
y_sort = y[index]
Pre = []
Rec = []
for i, item in enumerate(pred_sort):
if i == 0: # 因为计算precision的时候分母要用到i，当i为0时会出错，所以单独列出
Pre.append(1)
Rec.append(0)
else:
Pre.append(np.sum((y_sort[:i] == 1)) / i) # 从大到小阈值排列，所以全部都预测为正例
Rec.append(np.sum((y_sort[:i] == 1)) / pos)
Pre.append(0)
Rec.append(1) # 加入最后一个点，使图像封闭
pr_value = get_aupr(Rec, Pre)
# 画图
plt.figure('PR Curve')
plt.plot(Rec, Pre, 'r')
label = "AUPR: " + str(pr_value)
plt.title('PR Curve')
plt.plot([0, 1], [1, 0], 'k--', linewidth=0.8, label=label)
plt.xlim([-0.01, 1.01])
plt.ylim([-0.01, 1.01])
plt.ylabel('Precision')
plt.xlabel('Recall')
plt.legend(loc='lower right')
plt.show()
return pr_value
def ROC_curve(y, pred):
pos = np.sum(y == 1)
neg = np.sum(y == 0)
pred_sort = np.sort(pred)[::-1] # 从大到小排序
index = np.argsort(pred)[::-1] # 从大到小排序
y_sort = y[index]
tpr = []
fpr = []
thr = []
for i, item in enumerate(pred_sort):
tpr.append(np.sum((y_sort[:i] == 1)) / pos)
fpr.append(np.sum((y_sort[:i] == 0)) / neg)
thr.append(item)
tpr.append(1)
fpr.append(1)
auc_value = get_auroc(fpr, tpr)
# 画图
plt.figure('ROC Curve')
plt.title('ROC Curve')
plt.plot(fpr, tpr, 'r')
label = "AUROC: " + str(auc_value)
plt.plot([[0, 0], [1, 1]], 'k--', linewidth=0.8, label=label)
plt.xlim([-0.01, 1.01])
plt.ylim([-0.01, 01.01])
plt.ylabel('True Positive Rate')
plt.xlabel('False Positive Rate')
plt.legend(loc="lower right") # 若需显示label,必须在show之前加这一句
plt.show()
return auc_value
# 计算AUROC值
def get_auroc(fpr, tpr):
auc_value = 0.0
for ix in range(len(fpr[:-1])):
x_right, x_left = fpr[ix], fpr[ix + 1]
y_top, y_bottom = tpr[ix], tpr[ix + 1]
temp_area = abs(x_right - x_left) * (y_top + y_bottom) * 0.5
auc_value += temp_area
return auc_value
# 计算AUPR值
def get_aupr(rec, pre):
pr_value = 0.0
for ix in range(len(rec[:-1])):
x_right, x_left = rec[ix], rec[ix + 1]
y_top, y_bottom = pre[ix], pre[ix + 1]
temp_area = abs(x_right - x_left) * (y_top + y_bottom) * 0.5
pr_value += temp_area
return pr_value
def confusion_matrix(true_label, predict_label):
pos_label_count = 0
for k in range(len(true_label)):
if true_label[k] == 1:
pos_label_count += 1
pre_pos_count = 0
for k in range(len(predict_label)):
if predict_label[k] == 1:
pre_pos_count += 1
correct_predict = 0
for pre, tru in zip(predict_label, true_label):
if pre == 1 and tru == 1:
correct_predict += 1
TP = correct_predict
FN = pos_label_count - TP
FP = pre_pos_count - TP
TN = len(true_label) - pos_label_count - FP
precision = TP / (TP + FP)
recall = TP / (TP + FN)
tpr = recall
fpr = FP / (TN + FP)
F1 = 2 * precision * recall / (precision + recall)
return TP, FN, FP, TN, precision, recall, F1
if __name__ == '__main__':
io = '../Diabetes.xls'
dataframe = pd.read_excel(io)
dataset = dataframe.values
x = dataset[:, 1:]
y = dataset[:, 0]
splits = 10
probability_all = []
total_acc = 0
order_id = []
predict_all = []
sfolder = StratifiedKFold(n_splits=splits, shuffle=True) # 十折交叉验证划分数据集
for num, (train, test) in enumerate(sfolder.split(x, y)):
x_train = x[train, :]
y_train = y[train]
x_test = x[test, :]
y_test = y[test]
order_id.extend(test)
clt = tree.DecisionTreeClassifier(min_samples_leaf=10) # 决策树拟合训练数据
clt.fit(x_train, y_train)
y_pre = clt.predict(x_test)
y_pre_pro = clt.predict_proba(x_test)
predict_all.extend(y_pre)
acc = accuracy_score(y_test, y_pre)
total_acc += acc
print('第', num + 1, '折验证的准确率', acc)
for i in range(y_pre_pro.shape[0]):
probability_all.append(y_pre_pro[i][1])
print("十折交叉验证的平均准确率为：", total_acc / splits)
PRAUC = PR_curve(y[order_id], probability_all)
ROCAUC = ROC_curve(y[order_id], probability_all)
TP, FN, FP, TN, precision, recall, F1 = confusion_matrix(y[order_id], predict_all)
conf_matrix = np.array([TP, FN, FP, TN]).reshape((2, 2))
print('-------------------------------------------')
print('混淆矩阵：')
print(conf_matrix)
print('查准率P值：', precision)
print('查全率R值：', recall)
print('F1度量值：', F1)
print('PR曲线的AUC值为：', PRAUC)
print('AOC曲线的AUC值为：', ROCAUC)
