15次 手写数字识别-小数据集

补交作业：4.K均值算法 https://www.cnblogs.com/a188182/p/13057563.html（这次作业看错时间忘记交了，希望老师给个补交机会）

                  12.朴素贝叶斯-垃圾邮件分类 https://www.cnblogs.com/a188182/p/13060368.html（这次作业家里没电，没有交到，希望老师给个补交机会）     

1.手写数字数据集

• from sklearn.datasets import load_digits
• digits = load_digits()

1 from sklearn.datasets import load_digits
2 import numpy as np
3 
4 digits = load_digits()
5 x_data = digits.data.astype(np.float32)
6 y_data = digits.target.astype(np.float32).reshape(-1, 1)

2.图片数据预处理

• x：归一化MinMaxScaler()
• y：独热编码OneHotEncoder()或to_categorical
• 训练集测试集划分
• 张量结构

 1 from sklearn.model_selection import train_test_split
 2 from sklearn.preprocessing import MinMaxScaler, OneHotEncoder
 3 
 4 scaler = MinMaxScaler()
 5 x_data = scaler.fit_transform(x_data)
 6 print(x_data)
 7 x = x_data.reshape(-1, 8, 8, 1)  # 转换为图片格式
 8 y = OneHotEncoder().fit_transform(y_data).todense()
 9 # 训练集测试集划分
10 x_train, x_test, y_train, y_test = train_test_split(x, y, test_size=0.2, random_state=0, stratify=y)
11 print(x_train.shape, x_test.shape, y_train.shape, y_test.shape)

3.设计卷积神经网络结构

• 绘制模型结构图，并说明设计依据。

 1 from tensorflow.keras.models import Sequential
 2 from tensorflow.keras.layers import Dense,Dropout,Flatten,Conv2D,MaxPool2D
 3 import matplotlib.pyplot as plt
 4 # 3.设计卷积神经网络结构
 5 # 建立模型
 6 model = Sequential()
 7 ks = [3, 3]  # 卷积核
 8 # 一层卷积
 9 model.add(Conv2D(filters=16, kernel_size=ks, padding='same', input_shape=x_train.shape[1:], activation='relu'))
10 # 池化层
11 model.add(MaxPool2D(pool_size=(2, 2)))
12 model.add(Dropout(0.25))
13 # 二层卷积
14 model.add(Conv2D(filters=32, kernel_size=ks, padding='same', activation='relu'))
15 # 池化层
16 model.add(MaxPool2D(pool_size=(2, 2)))
17 model.add(Dropout(0.25))
18 # 三层卷积
19 model.add(Conv2D(filters=64, kernel_size=ks, padding='same', activation='relu'))
20 # 四层卷积
21 model.add(Conv2D(filters=128, kernel_size=ks, padding='same', activation='relu'))
22 # 池化层
23 model.add(MaxPool2D(pool_size=(2, 2)))
24 model.add(Dropout(0.25))
25 # 平坦层
26 model.add(Flatten())
27 # 全连接层
28 model.add(Dense(128, activation='relu'))
29 model.add(Dropout(0.25))
30 # 激活函数
31 model.add(Dense(10, activation='softmax'))
32 model.summary()

4.模型训练

 1 #绘制模型结构图
 2 def show_train_history(train_history, train, validation):
 3     plt.plot(train_history.history[train])
 4     plt.plot(train_history.history[validation])
 5     plt.title('Train History')
 6     plt.ylabel('train')
 7     plt.xlabel('epoch')
 8     plt.legend(['train', 'validation'], loc='upper left')
 9     plt.show()
10 model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
11 train_history = model.fit(x=x_train, y=y_train, validation_split=0.2, batch_size=300, epochs=10, verbose=2)
12 # 准确率
13 show_train_history(train_history, 'accuracy', 'val_accuracy')
14 # 损失率
15 show_train_history(train_history, 'loss', 'val_loss')

准确率：

损失率：

5.模型评价

• model.evaluate()
• 交叉表与交叉矩阵
• pandas.crosstab
• seaborn.heatmap

 1 import pandas as pd
 2 import seaborn as sns
 3 # 5、模型评价
 4 #模型评估
 5 score = model.evaluate(x_test, y_test)[1]
 6 print('模型准确率=',score)
 7 # 预测值
 8 y_pre = model.predict_classes(x_test)
 9 y_pre[:10]
10 y_test1 = np.argmax(y_test, axis=1).reshape(-1)
11 y_true = np.array(y_test1)[0]
12 y_true.shape
13 pd.crosstab(y_true, y_pre, rownames=['true'], colnames=['predict'])
14 # 交叉矩阵
15 y_test1 = y_test1.tolist()[0]
16 a = pd.crosstab(np.array(y_test1), y_pre, rownames=['Lables'], colnames=['predict'])
17 df = pd.DataFrame(a)
18 print(df)
19 sns.heatmap(df, annot=True, cmap="Reds", linewidths=0.2, linecolor='G')
20 plt.show()