如何使用sklearn的评估函数来完成keras模型的评估

在我使用keras做模型的时候，需要评估其模型的召回率、精度、F1-score。在百度搜索自定义评估函数的时候，发现自己代码功底太弱，就想把sklearn的classification_report()评估函数拿来用，发现其存在的问题，并解决。废话不多说，直接上代码。
1、多分类任务中 softmax的值是小数，第一步需要将其转换成onehot
2、通过sklearn封装的函数将onehot来转换成整数标签（不要忘记y_test的标签在喂入模型中是先进行onehot的，所以一定要在之后转换成整数标签值）

import warnings
warnings.filterwarnings("ignore")
import pandas as pd
import seaborn as sns
sns.set(style="white", color_codes=True)
import matplotlib.pyplot as plt
import numpy as np
import os
from keras.models import Sequential, Model
from keras.layers import Input, merge
from keras.layers.core import Dense, Dropout, Activation, Flatten, Lambda, Reshape, Permute
from keras.layers.convolutional import Conv2D, MaxPooling2D, Conv3D, MaxPooling3D, Conv1D, MaxPooling1D
from keras.layers.recurrent import GRU, LSTM
from keras.layers.advanced_activations import ELU, LeakyReLU
from keras.layers.normalization import BatchNormalization
from keras.utils import np_utils
from keras.optimizers import SGD, Adam, Adagrad, Adadelta, RMSprop
from keras.callbacks import EarlyStopping, ModelCheckpoint
from keras.constraints import max_norm
from keras import backend as K
from imblearn.under_sampling import RandomUnderSampler
from collections import Counter
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import Imputer, LabelEncoder, OneHotEncoder
from sklearn.metrics import classification_report, accuracy_score

    # sortmax 结果转 onehot
    def props_to_onehot(props):
        if isinstance(props, list):
            props = np.array(props)
        a = np.argmax(props, axis=1)
        b = np.zeros((len(a), props.shape[1]))
        b[np.arange(len(a)), a] = 1
        return b

    #将标签 onehot
    y_train = np_utils.to_categorical(y_train)
    y_test = np_utils.to_categorical(y_test)


    classifier = Sequential()
    classifier.add(Dense(input_dim=15, units=256, activation='relu'))
    classifier.add(Dense(units=30, activation='relu'))
    classifier.add(Dense(units=11, activation='softmax'))
    adam = Adam(lr=5e-4, beta_1=0.9, beta_2=0.999, epsilon=1e-08)
    classifier.compile(loss='categorical_crossentropy',
                       optimizer=adam,
                       metrics=['accuracy'])
    classifier.fit(X_train, y_train, batch_size=32, epochs=50, verbose=2)
    classifier.save_weights('./model.h5')
    y_pred = classifier.predict(X_test)
    loss, acc_ANN = classifier.evaluate(X_test, y_test, verbose=2)

    #onehot转换整数标签
    enc = OneHotEncoder(categories='auto')
    #假设样本有11类
    enc.fit([[0], [1], [2], [3], [4], [5], [6], [7], [8], [9], [10]])
    y_pred = enc.inverse_transform(props_to_onehot(y_pred))
    y_test = enc.inverse_transform(props_to_onehot(y_test))

    report_ANN = classification_report(y_test, y_pred)
    print('acc_ANN is', acc_ANN)
    print('report_ANN', report_ANN)