tensorflow 1-1,結構化數據建模流程範例

時間 2021-08-14

標籤 json session ide svg 函數測試編碼 lua orm 繼承欄目 JavaScript 简体版

原文原文鏈接

tensorflow 1-1,結構化數據建模流程範例

1-1,結構化數據建模流程範例

一，準備數據

titanic數據集的目標是根據乘客信息預測他們在Titanic號撞擊冰山沉沒後可否生存。json

結構化數據通常會使用Pandas中的DataFrame進行預處理。session

import numpy as np 
import pandas as pd 
import matplotlib.pyplot as plt
import tensorflow as tf 
from tensorflow.keras import models,layers

dftrain_raw = pd.read_csv('./data/titanic/train.csv')
dftest_raw = pd.read_csv('./data/titanic/test.csv')
dftrain_raw.head(10)

字段說明：ide

Survived:0表明死亡，1表明存活【y標籤】
Pclass:乘客所持票類，有三種值(1,2,3) 【轉換成onehot編碼】
Name:乘客姓名【捨去】
Sex:乘客性別【轉換成bool特徵】
Age:乘客年齡(有缺失) 【數值特徵，添加「年齡是否缺失」做爲輔助特徵】
SibSp:乘客兄弟姐妹/配偶的個數(整數值) 【數值特徵】
Parch:乘客父母/孩子的個數(整數值)【數值特徵】
Ticket:票號(字符串)【捨去】
Fare:乘客所持票的價格(浮點數，0-500不等) 【數值特徵】
Cabin:乘客所在船艙(有缺失) 【添加「所在船艙是否缺失」做爲輔助特徵】
Embarked:乘客登船港口:S、C、Q(有缺失)【轉換成onehot編碼，四維度 S,C,Q,nan】

利用Pandas的數據可視化功能咱們能夠簡單地進行探索性數據分析EDA（Exploratory Data Analysis）。svg

label分佈狀況函數

%matplotlib inline
%config InlineBackend.figure_format = 'png'
ax = dftrain_raw['Survived'].value_counts().plot(kind = 'bar',
     figsize = (12,8),fontsize=15,rot = 0)
ax.set_ylabel('Counts',fontsize = 15)
ax.set_xlabel('Survived',fontsize = 15)
plt.show()

年齡分佈狀況測試

%matplotlib inline
%config InlineBackend.figure_format = 'png'
ax = dftrain_raw['Age'].plot(kind = 'hist',bins = 20,color= 'purple',
                    figsize = (12,8),fontsize=15)

ax.set_ylabel('Frequency',fontsize = 15)
ax.set_xlabel('Age',fontsize = 15)
plt.show()

年齡和label的相關性編碼

%matplotlib inline
%config InlineBackend.figure_format = 'png'
ax = dftrain_raw.query('Survived == 0')['Age'].plot(kind = 'density',
                      figsize = (12,8),fontsize=15)
dftrain_raw.query('Survived == 1')['Age'].plot(kind = 'density',
                      figsize = (12,8),fontsize=15)
ax.legend(['Survived==0','Survived==1'],fontsize = 12)
ax.set_ylabel('Density',fontsize = 15)
ax.set_xlabel('Age',fontsize = 15)
plt.show()

下面爲正式的數據預處理lua

def preprocessing(dfdata):

    dfresult= pd.DataFrame()

    #Pclass
    dfPclass = pd.get_dummies(dfdata['Pclass'])
    dfPclass.columns = ['Pclass_' +str(x) for x in dfPclass.columns ]
    dfresult = pd.concat([dfresult,dfPclass],axis = 1)

    #Sex
    dfSex = pd.get_dummies(dfdata['Sex'])
    dfresult = pd.concat([dfresult,dfSex],axis = 1)

    #Age
    dfresult['Age'] = dfdata['Age'].fillna(0)
    dfresult['Age_null'] = pd.isna(dfdata['Age']).astype('int32')

    #SibSp,Parch,Fare
    dfresult['SibSp'] = dfdata['SibSp']
    dfresult['Parch'] = dfdata['Parch']
    dfresult['Fare'] = dfdata['Fare']

    #Carbin
    dfresult['Cabin_null'] =  pd.isna(dfdata['Cabin']).astype('int32')

    #Embarked
    dfEmbarked = pd.get_dummies(dfdata['Embarked'],dummy_na=True)
    dfEmbarked.columns = ['Embarked_' + str(x) for x in dfEmbarked.columns]
    dfresult = pd.concat([dfresult,dfEmbarked],axis = 1)

    return(dfresult)

x_train = preprocessing(dftrain_raw)
y_train = dftrain_raw['Survived'].values

x_test = preprocessing(dftest_raw)
y_test = dftest_raw['Survived'].values

print("x_train.shape =", x_train.shape )
print("x_test.shape =", x_test.shape )

x_train.shape = (712, 15)
x_test.shape = (179, 15)

二，定義模型

使用Keras接口有如下3種方式構建模型：使用Sequential按層順序構建模型，使用函數式API構建任意結構模型，繼承Model基類構建自定義模型。orm

此處選擇使用最簡單的Sequential，按層順序模型。繼承

tf.keras.backend.clear_session()

model = models.Sequential()
model.add(layers.Dense(20,activation = 'relu',input_shape=(15,)))
model.add(layers.Dense(10,activation = 'relu' ))
model.add(layers.Dense(1,activation = 'sigmoid' ))

model.summary()

Model: "sequential"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense (Dense)                (None, 20)                320       
_________________________________________________________________
dense_1 (Dense)              (None, 10)                210       
_________________________________________________________________
dense_2 (Dense)              (None, 1)                 11        
=================================================================
Total params: 541
Trainable params: 541
Non-trainable params: 0
_________________________________________________________________

三，訓練模型

訓練模型一般有3種方法，內置fit方法，內置train_on_batch方法，以及自定義訓練循環。此處咱們選擇最經常使用也最簡單的內置fit方法。

# 二分類問題選擇二元交叉熵損失函數
model.compile(optimizer='adam',
            loss='binary_crossentropy',
            metrics=['AUC'])

history = model.fit(x_train,y_train,
                    batch_size= 64,
                    epochs= 30,
                    validation_split=0.2 #分割一部分訓練數據用於驗證
                   )

Train on 569 samples, validate on 143 samples
Epoch 1/30
569/569 [==============================] - 1s 2ms/sample - loss: 3.5841 - AUC: 0.4079 - val_loss: 3.4429 - val_AUC: 0.4129
Epoch 2/30
569/569 [==============================] - 0s 102us/sample - loss: 2.6093 - AUC: 0.3967 - val_loss: 2.4886 - val_AUC: 0.4139
Epoch 3/30
569/569 [==============================] - 0s 68us/sample - loss: 1.8375 - AUC: 0.4003 - val_loss: 1.7383 - val_AUC: 0.4223
Epoch 4/30
569/569 [==============================] - 0s 83us/sample - loss: 1.2545 - AUC: 0.4390 - val_loss: 1.1936 - val_AUC: 0.4765
Epoch 5/30
569/569 [==============================] - ETA: 0s - loss: 1.4435 - AUC: 0.375 - 0s 90us/sample - loss: 0.9141 - AUC: 0.5192 - val_loss: 0.8274 - val_AUC: 0.5584
Epoch 6/30
569/569 [==============================] - 0s 110us/sample - loss: 0.7052 - AUC: 0.6290 - val_loss: 0.6596 - val_AUC: 0.6880
Epoch 7/30
569/569 [==============================] - 0s 90us/sample - loss: 0.6410 - AUC: 0.7086 - val_loss: 0.6519 - val_AUC: 0.6845
Epoch 8/30
569/569 [==============================] - 0s 93us/sample - loss: 0.6246 - AUC: 0.7080 - val_loss: 0.6480 - val_AUC: 0.6846
Epoch 9/30
569/569 [==============================] - 0s 73us/sample - loss: 0.6088 - AUC: 0.7113 - val_loss: 0.6497 - val_AUC: 0.6838
Epoch 10/30
569/569 [==============================] - 0s 79us/sample - loss: 0.6051 - AUC: 0.7117 - val_loss: 0.6454 - val_AUC: 0.6873
Epoch 11/30
569/569 [==============================] - 0s 96us/sample - loss: 0.5972 - AUC: 0.7218 - val_loss: 0.6369 - val_AUC: 0.6888
Epoch 12/30
569/569 [==============================] - 0s 92us/sample - loss: 0.5918 - AUC: 0.7294 - val_loss: 0.6330 - val_AUC: 0.6908
Epoch 13/30
569/569 [==============================] - 0s 75us/sample - loss: 0.5864 - AUC: 0.7363 - val_loss: 0.6281 - val_AUC: 0.6948
Epoch 14/30
569/569 [==============================] - 0s 104us/sample - loss: 0.5832 - AUC: 0.7426 - val_loss: 0.6240 - val_AUC: 0.7030
Epoch 15/30
569/569 [==============================] - 0s 74us/sample - loss: 0.5777 - AUC: 0.7507 - val_loss: 0.6200 - val_AUC: 0.7066
Epoch 16/30
569/569 [==============================] - 0s 79us/sample - loss: 0.5726 - AUC: 0.7569 - val_loss: 0.6155 - val_AUC: 0.7132
Epoch 17/30
569/569 [==============================] - 0s 99us/sample - loss: 0.5674 - AUC: 0.7643 - val_loss: 0.6070 - val_AUC: 0.7255
Epoch 18/30
569/569 [==============================] - 0s 97us/sample - loss: 0.5631 - AUC: 0.7721 - val_loss: 0.6061 - val_AUC: 0.7305
Epoch 19/30
569/569 [==============================] - 0s 73us/sample - loss: 0.5580 - AUC: 0.7792 - val_loss: 0.6027 - val_AUC: 0.7332
Epoch 20/30
569/569 [==============================] - 0s 85us/sample - loss: 0.5533 - AUC: 0.7861 - val_loss: 0.5997 - val_AUC: 0.7366
Epoch 21/30
569/569 [==============================] - 0s 87us/sample - loss: 0.5497 - AUC: 0.7926 - val_loss: 0.5961 - val_AUC: 0.7433
Epoch 22/30
569/569 [==============================] - 0s 101us/sample - loss: 0.5454 - AUC: 0.7987 - val_loss: 0.5943 - val_AUC: 0.7438
Epoch 23/30
569/569 [==============================] - 0s 100us/sample - loss: 0.5398 - AUC: 0.8057 - val_loss: 0.5926 - val_AUC: 0.7492
Epoch 24/30
569/569 [==============================] - 0s 79us/sample - loss: 0.5328 - AUC: 0.8122 - val_loss: 0.5912 - val_AUC: 0.7493
Epoch 25/30
569/569 [==============================] - 0s 86us/sample - loss: 0.5283 - AUC: 0.8147 - val_loss: 0.5902 - val_AUC: 0.7509
Epoch 26/30
569/569 [==============================] - 0s 67us/sample - loss: 0.5246 - AUC: 0.8196 - val_loss: 0.5845 - val_AUC: 0.7552
Epoch 27/30
569/569 [==============================] - 0s 72us/sample - loss: 0.5205 - AUC: 0.8271 - val_loss: 0.5837 - val_AUC: 0.7584
Epoch 28/30
569/569 [==============================] - 0s 74us/sample - loss: 0.5144 - AUC: 0.8302 - val_loss: 0.5848 - val_AUC: 0.7561
Epoch 29/30
569/569 [==============================] - 0s 77us/sample - loss: 0.5099 - AUC: 0.8326 - val_loss: 0.5809 - val_AUC: 0.7583
Epoch 30/30
569/569 [==============================] - 0s 80us/sample - loss: 0.5071 - AUC: 0.8349 - val_loss: 0.5816 - val_AUC: 0.7605

四，評估模型

咱們首先評估一下模型在訓練集和驗證集上的效果。

%matplotlib inline
%config InlineBackend.figure_format = 'svg'

import matplotlib.pyplot as plt

def plot_metric(history, metric):
    train_metrics = history.history[metric]
    val_metrics = history.history['val_'+metric]
    epochs = range(1, len(train_metrics) + 1)
    plt.plot(epochs, train_metrics, 'bo--')
    plt.plot(epochs, val_metrics, 'ro-')
    plt.title('Training and validation '+ metric)
    plt.xlabel("Epochs")
    plt.ylabel(metric)
    plt.legend(["train_"+metric, 'val_'+metric])
    plt.show()

plot_metric(history,"loss")

plot_metric(history,"AUC")

咱們再看一下模型在測試集上的效果.

model.evaluate(x = x_test,y = y_test)

[0.5191367897907448, 0.8122605]

五，使用模型

#預測機率
model.predict(x_test[0:10])
#model(tf.constant(x_test[0:10].values,dtype = tf.float32)) #等價寫法

array([[0.26501188],
       [0.40970832],
       [0.44285864],
       [0.78408605],
       [0.47650957],
       [0.43849158],
       [0.27426785],
       [0.5962582 ],
       [0.59476686],
       [0.17882936]], dtype=float32)

#預測類別
model.predict_classes(x_test[0:10])

array([[0],
       [0],
       [0],
       [1],
       [0],
       [0],
       [0],
       [1],
       [1],
       [0]], dtype=int32)

六，保存模型

能夠使用Keras方式保存模型，也能夠使用TensorFlow原生方式保存。前者僅僅適合使用Python環境恢復模型，後者則能夠跨平臺進行模型部署。

推薦使用後一種方式進行保存。

1，Keras方式保存

# 保存模型結構及權重

model.save('./data/keras_model.h5')  

del model  #刪除現有模型

# identical to the previous one
model = models.load_model('./data/keras_model.h5')
model.evaluate(x_test,y_test)

[0.5191367897907448, 0.8122605]

# 保存模型結構
json_str = model.to_json()

# 恢復模型結構
model_json = models.model_from_json(json_str)

#保存模型權重
model.save_weights('./data/keras_model_weight.h5')

# 恢復模型結構
model_json = models.model_from_json(json_str)
model_json.compile(
        optimizer='adam',
        loss='binary_crossentropy',
        metrics=['AUC']
    )

# 加載權重
model_json.load_weights('./data/keras_model_weight.h5')
model_json.evaluate(x_test,y_test)

[0.5191367897907448, 0.8122605]

2，TensorFlow原生方式保存

# 保存權重，該方式僅僅保存權重張量
model.save_weights('./data/tf_model_weights.ckpt',save_format = "tf")

# 保存模型結構與模型參數到文件,該方式保存的模型具備跨平臺性便於部署

model.save('./data/tf_model_savedmodel', save_format="tf")
print('export saved model.')

model_loaded = tf.keras.models.load_model('./data/tf_model_savedmodel')
model_loaded.evaluate(x_test,y_test)