什么是K折交叉验证？
Keras实现K折交叉验证
# -*- coding: utf-8 -*- # model_train.py # Python 3.6.8, TensorFlow 2.3.0, Keras 2.4.3 # 导入模块 import keras as K import pandas as pd from sklearn.model_selection import KFold # 读取CSV数据集 # 该函数的传入参数为csv_file_path: csv文件路径 def load_data(sv_file_path): iris = pd.read_csv(sv_file_path) target_var = 'class' # 目标变量 # 数据集的特征 features = list(iris.columns) features.remove(target_var) # 目标变量的类别 Class = iris[target_var].unique() # 目标变量的类别字典 Class_dict = dict(zip(Class, range(len(Class)))) # 增加一列target, 将目标变量转化为类别变量 iris['target'] = iris[target_var].apply(lambda x: Class_dict[x]) return features, 'target', iris # 创建模型 def create_model(): init = K.initializers.glorot_uniform(seed=1) simple_adam = K.optimizers.Adam() model = K.models.Sequential() model.add(K.layers.Dense(units=5, input_dim=4, kernel_initializer=init, activation='relu')) model.add(K.layers.Dense(units=6, kernel_initializer=init, activation='relu')) model.add(K.layers.Dense(units=3, kernel_initializer=init, activation='softmax')) model.compile(loss='sparse_categorical_crossentropy', optimizer=simple_adam, metrics=['accuracy']) return model def main(): # 1. 读取CSV数据集 print("Loading Iris data into memory") n_split = 10 features, target, data = load_data("./iris_data.csv") x = data[features] y = data[target] avg_accuracy = 0 avg_loss = 0 for train_index, test_index in KFold(n_split).split(x): print("test index: ", test_index) x_train, x_test = x.iloc[train_index], x.iloc[test_index] y_train, y_test = y.iloc[train_index], y.iloc[test_index] print("create model and train model") model = create_model() model.fit(x_train, y_train, batch_size=1, epochs=80, verbose=0) print('Model evaluation: ', model.evaluate(x_test, y_test)) avg_accuracy += model.evaluate(x_test, y_test)[1] avg_loss += model.evaluate(x_test, y_test)[0] print("K fold average accuracy: {}".format(avg_accuracy / n_split)) print("K fold average accuracy: {}".format(avg_loss / n_split)) main() 模型的输出结果如下：
| Iteration | loss | accuracy |
|---|---|---|
| 1 | 0.00056 | 1.0 |
| 2 | 0.00021 | 1.0 |
| 3 | 0.00022 | 1.0 |
| 4 | 0.00608 | 1.0 |
| 5 | 0.21925 | 0.8667 |
| 6 | 0.52390 | 0.8667 |
| 7 | 0.00998 | 1.0 |
| 8 | 0.04431 | 1.0 |
| 9 | 0.14590 | 1.0 |
| 10 | 0.21286 | 0.8667 |
| avg | 0.11633 | 0.9600 |
10折交叉验证的平均loss为0.11633,平均准确率为96.00%。
总结
版权声明：本文内容由互联网用户自发贡献，该文观点仅代表作者本人。本站仅提供信息存储空间服务，不拥有所有权，不承担相关法律责任。如发现本站有涉嫌侵权/违法违规的内容， 请联系我们举报，一经查实，本站将立刻删除。
发布者：全栈程序员-站长，转载请注明出处：https://javaforall.net/233872.html原文链接：https://javaforall.net
