机器学习——支持向量机

1 from sklearn.datasets import load_digits
2 from sklearn.metrics import classification_report
3 from sklearn.model_selection import train_test_split
4 from sklearn.preprocessing import StandardScaler
5 from sklearn.svm import LinearSVC
6 import numpy as np
7 import matplotlib.pyplot as plt

 1 def hw_recognition():
 2     digits = load_digits()
 3 
 4     X_train, X_test, Y_train, Y_test = train_test_split(digits.data, digits.target, test_size=0.25, random_state=23)
 5 
 6     ss = StandardScaler()
 7     # fit 是实例方法，必须由实例调用
 8     X_train = ss.fit_transform(X_train)
 9     X_test = ss.transform(X_test)
10 
11     lsvc = LinearSVC()
12     lsvc.fit(X_train, Y_train)
13 
14     # 预测测试集数据
15     # Y_predict = lsvc.predict(np.array([X_test[90]]))
16     # print(Y_predict)
17 
18     # 生成评估报告
19     # 精确率precision = 正正 / (正正 + 反正)
20     # 按列计算
21     #
22     # 召回率recall = 正正 / (正正 + 正反)
23     # 按行计算
24     # 按行解释正反：正正-真正的正例预测为正例；正反-真正的正例预测为反例
25     #            反正-真正的反例预测为正例；反正-真正的反例预测为正例
26     Y_predict = lsvc.predict(X_test)
27     print(classification_report(Y_test, Y_predict, target_names=digits.target_names.astype(str)))
28 
29     # 对比预测的数字与原图像中的数字
30     # Y_predict = lsvc.predict(np.array([digits.data[307]]))
31     # plt.imshow(digits.images[307])
32     # print(Y_predict)
33     # plt.show()

 1 真实数据       预测数据
 2 ---------------------------
 3            18正     12反   
 4  20正      12正正    8正反  
 5  10反       6反正    4反反  
 6 ---------------------------
 7 
 8 精确率precision = 正正/(正正+反正)            按列计算
 9 
10 召回率recall = 正正/(正正+正反)            按行计算

精确率      召回率
             precision    recall  f1-score   support

          0       1.00      1.00      1.00        37
          1       0.86      0.90      0.88        48
          2       1.00      0.98      0.99        46
          3       0.91      0.98      0.94        41
          4       1.00      0.96      0.98        49
          5       0.91      0.98      0.94        50
          6       0.98      0.98      0.98        41
          7       0.92      0.95      0.93        37
          8       0.93      0.89      0.91        46
          9       1.00      0.91      0.95        55

avg / total       0.95      0.95      0.95       450