手寫數字識別-小數據集

作業補交說明

第五次作業：https://www.cnblogs.com/lzhdonald/p/12758516.html

當時已經寫完了作業存在草稿，忘記發佈了沒有提交，第二天發現不能提交了，當時也有和老師說明了情況。

1.手寫數字數據集

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

#導入手寫數字數據集
from sklearn.datasets import load_digits
import numpy as np
digits = load_digits()

2.圖片數據預處理

• x：歸一化MinMaxScaler()
• y：獨熱編碼OneHotEncoder()或to_categorical
• 訓練集測試集劃分
• 張量結構

# 歸一化MinMaxScaler()
from sklearn.preprocessing import MinMaxScaler
X_data = digits.data.astype(np.float32)
scaler = MinMaxScaler()
X_data = scaler.fit_transform(X_data)
print("歸一化後",X_data)
# 轉化爲圖片的格式
X=X_data.reshape(-1,8,8,1)

獨熱編碼：

# 獨熱編碼
from sklearn.preprocessing import OneHotEncoder
# y = digits.target.reshape(-1,1)
#將Y_data變爲一列
y = digits.target.astype(np.float32).reshape(-1,1)  
Y = OneHotEncoder().fit_transform(y).todense() #張量結構todense
print("獨熱編碼：",Y)
# 切分數據集
from sklearn.model_selection import train_test_split
X_train,X_test,y_train,y_test = train_test_split(X,Y,test_size=0.2,random_state=0,stratify=Y)
print(X_train,X_test,y_train,y_test)
print("X_data.shape:",X_data.shape)
print("X.shape",X.shape)

3.設計卷積神經網絡結構

• 繪製模型結構圖，並說明設計依據。

from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense,Dropout,Conv2D,MaxPool2D,Flatten
#3、建立模型
model = Sequential()
ks = (3, 3)  # 卷積核的大小
input_shape = X_train.shape[1:]
# 一層卷積，padding='same',tensorflow會對輸入自動補0
model.add(Conv2D(filters=16, kernel_size=ks, padding='same', input_shape=input_shape, activation='relu'))
# 池化層1
model.add(MaxPool2D(pool_size=(2, 2)))
# 防止過擬合，隨機丟掉連接
model.add(Dropout(0.25))
# 二層卷積
model.add(Conv2D(filters=32, kernel_size=ks, padding='same', activation='relu'))
# 池化層2
model.add(MaxPool2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
# 三層卷積
model.add(Conv2D(filters=64, kernel_size=ks, padding='same', activation='relu'))
# 四層卷積
model.add(Conv2D(filters=128, kernel_size=ks, padding='same', activation='relu'))
# 池化層3
model.add(MaxPool2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
# 平坦層
model.add(Flatten())
# 全連接層
model.add(Dense(128, activation='relu'))
model.add(Dropout(0.25))
# 激活函數softmax
model.add(Dense(10, activation='softmax'))
print(model.summary())

4.模型訓練

import matplotlib.pyplot as plt
# 畫圖
def show_train_history(train_history, train, validation):
    plt.plot(train_history.history[train])
    plt.plot(train_history.history[validation])
    plt.title('Train History')
    plt.ylabel('train')
    plt.xlabel('epoch')
    plt.legend(['train', 'validation'], loc='upper left')
    plt.show()

# 4、模型訓練
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
train_history = model.fit(x=X_train, y=y_train, validation_split=0.2, batch_size=300, epochs=10, verbose=2)
# 準確率
show_train_history(train_history, 'accuracy', 'val_accuracy')
# 損失率
show_train_history(train_history, 'loss', 'val_loss')

5.模型評價

• model.evaluate()
• 交叉表與交叉矩陣
• pandas.crosstab
• seaborn.heatmap

import pandas as pd
import seaborn as sns
# model.evaluate()
score = model.evaluate(X_test, y_test)
print('score：', score)
# 預測值
y_pred = model.predict_classes(X_test)
print('y_pred：', y_pred[:10])
# 交叉表與交叉矩陣
y_test1 = np.argmax(y_test, axis=1).reshape(-1)
y_true = np.array(y_test1)[0]
# 交叉表查看預測數據與原數據對比
# pandas.crosstab
pd.crosstab(y_true, y_pred, rownames=['true'], colnames=['predict'])
# 交叉矩陣
# seaborn.heatmap
y_test1 = y_test1.tolist()[0]
a = pd.crosstab(np.array(y_test1), y_pred, rownames=['Lables'], colnames=['Predict'])
# 轉換成屬dataframe
df = pd.DataFrame(a)
sns.heatmap(df, annot=True, cmap="Reds", linewidths=0.2, linecolor='G')
plt.show()