sklearn指标评估的python计算方法

# coding=utf-8
# 朴素贝叶斯方法
import sklearn
import numpy as np
from sklearn.model_selection import train_test_split
import matplotlib as mpl
import matplotlib.pyplot as plt
from sklearn.metrics import f1_score, precision_score, recall_score, auc,accuracy_score,roc_auc_score
from sklearn.metrics import zero_one_loss,log_loss

from sklearn.naive_bayes import GaussianNB
def plot_roc(name, labels, predictions, **kwargs):
  fp, tp, _ = sklearn.metrics.roc_curve(labels, predictions, pos_label=0)

  plt.plot(100*fp, 100*tp, label=name, linewidth=2, **kwargs)
  plt.xlabel('False positives [%]')
  plt.ylabel('True positives [%]')
  plt.xlim([-0.5,100.5])
  plt.ylim([-0.5,100.5])
  plt.grid(True)
  ax = plt.gca()
  ax.set_aspect('equal')
  plt.legend(loc='lower right')
  plt.savefig("./img/binary_roc.png")
  return auc(fp,tp)

def PrepareData():
    #制作数据集
    # binary
    path = '../dataset/dataset_supplement.npz'
    f = np.load(path)
    x, y = f['x'][:,:-1], f['x'][:,-1]
    f.close()
    print(np.array(x).shape)
    y = y.astype(np.int16)
    # print(np.bincount(y, weights=None, minlength=0))
    # y = make_label(y)
    # print(np.bincount(y, weights=None, minlength=0))
    # np.savez('./dataset/dataset_two_supplement.npz',x=x,y=y)

    # multi
    # path = '../dataset/dataset_two_supplement.npz'
    # f = np.load(path)
    # x, y = f['x'], f['y']
    # f.close()
    # print(np.array(x).shape)
    # print(np.bincount(y, weights=None, minlength=0))
    #制作训练集、验证集和训练集
    #比例：60%：20%：20%
    x_train, x_test, y_train, y_test = train_test_split(x, y, 
                                                        test_size=0.2,
                                                        shuffle = True, 
                                                        stratify = y    #根据标签y的比例分割
                                                        )
    x_train, x_valid, y_train, y_valid = train_test_split(
                                                        x_train, y_train, 
                                                        test_size=0.25,
                                                        shuffle = True, 
                                                        stratify = y_train
                                                        ) 
    #y_train = tf.one_hot(y_train, 2)
    return  x_train,y_train,x_valid,y_valid,x_test,y_test

x_train,y_train,x_valid,y_valid,x_test,y_test=PrepareData()  
model = GaussianNB()
model = model.fit(x_train, y_train)

# binary
# test_predictions = model.predict(x_test)
# y_true=y_test.astype('uint8')
# test_scores = 1-(test_predictions - test_predictions.min())/(test_predictions.max() - test_predictions.min())

# mpl.rcParams['figure.figsize'] = (12, 10)
# colors = plt.rcParams['axes.prop_cycle'].by_key()['color']
# auc = plot_roc("My Model", y_true, test_scores, color=colors[0])
# loss = zero_one_loss(y_true,test_predictions.round())
# acc = accuracy_score(y_true,test_predictions.round())
# precision = precision_score(y_true,test_predictions.round())
# print('acc: ',acc)
# print('loss: ',loss)
# print('precision: ',precision)
# print('auc: ',auc)

# multi
y_pred = model.predict_proba(x_test)
test_predictions = np.argmax(y_pred, axis=1)
y_true=y_test.astype('uint8')
test_scores = 1-(test_predictions - test_predictions.min())/(test_predictions.max() - test_predictions.min())
mpl.rcParams['figure.figsize'] = (12, 10)
colors = plt.rcParams['axes.prop_cycle'].by_key()['color']
auc = plot_roc("My Model", y_true, test_scores, color=colors[0])
loss = zero_one_loss(y_true,test_predictions)
precision = precision_score(y_true,test_predictions,average='weighted')
auc = roc_auc_score(y_true,y_pred,multi_class='ovr')

print('loss: ',loss)
print('precision: ',precision)
print('auc: ',auc)

from sklearn import metrics
print(metrics.classification_report(y_true, test_predictions,digits=4))

使用x_scores的方法进行计算评估指标的方法不准确，只有precision和accuracy是比较准确的，推荐使用classification_report的方法计算，AUC推荐plot_roc中的方法