Chapter 13

第三章模型搭建和评估

NotebookPython 379 cells

第三章模型搭建和评估

经过前面的探索性数据分析我们可以很清楚的了解到数据集的情况

In [1]python · cell 3

python

import pandas as pd
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt
from IPython.display import Image

In [2]python · cell 4

python

%matplotlib inline

In [3]python · cell 5

python

plt.rcParams['font.sans-serif'] = ['SimHei']  # 用来正常显示中文标签
plt.rcParams['axes.unicode_minus'] = False  # 用来正常显示负号
plt.rcParams['figure.figsize'] = (10, 6)  # 设置输出图片大小

In [4]python · cell 6

python

# 读取训练数集
train = pd.read_csv('train.csv')
train.shape

Output

(891, 12)

In [5]python · cell 7

python

train.head()

Output

   PassengerId  Survived  Pclass  \
0            1         0       3   
1            2         1       1   
2            3         1       3   
3            4         1       1   
4            5         0       3   

                                                Name     Sex   Age  SibSp  \
0                            Braund, Mr. Owen Harris    male  22.0      1   
1  Cumings, Mrs. John Bradley (Florence Briggs Th...  female  38.0      1   
2                             Heikkinen, Miss. Laina  female  26.0      0   
3       Futrelle, Mrs. Jacques Heath (Lily May Peel)  female  35.0      1   
4                           Allen, Mr. William Henry    male  35.0      0   

   Parch            Ticket     Fare Cabin Embarked  
0      0         A/5 21171   7.2500   NaN        S  
1      0          PC 17599  71.2833   C85        C  
2      0  STON/O2. 3101282   7.9250   NaN        S  
3      0            113803  53.1000  C123        S  
4      0            373450   8.0500   NaN        S

.dataframe tbody tr th:only-of-type { vertical-align: middle; } .dataframe tbody tr th { vertical-align: top; } .dataframe thead th { text-align: right; }

	PassengerId	Survived	Pclass	Name	Sex	Age	SibSp	Ticket	Fare	Cabin	Embarked
0	1	0	3	Braund, Mr. Owen Harris	male	22.0	1	A/5 21171	7.2500	NaN	S
1	2	1	1	Cumings, Mrs. John Bradley (Florence Briggs Th...	female	38.0	1	PC 17599	71.2833	C85	C
2	3	1	3	Heikkinen, Miss. Laina	female	26.0	0	STON/O2. 3101282	7.9250	NaN	S
3	4	1	1	Futrelle, Mrs. Jacques Heath (Lily May Peel)	female	35.0	1	113803	53.1000	C123	S
4	5	0	3	Allen, Mr. William Henry	male	35.0	0	373450	8.0500	NaN	S

特征工程

任务一：缺失值填充

对分类变量缺失值：填充某个缺失值字符(NA)、用最多类别的进行填充
对连续变量缺失值：填充均值、中位数、众数

In [6]python · cell 10

python

# 对分类变量进行填充
train['Cabin'] = train['Cabin'].fillna('NA')
train['Embarked'] = train['Embarked'].fillna('S')

In [7]python · cell 11

python

# 对连续变量进行填充
train['Age'] = train['Age'].fillna(train['Age'].mean())

In [8]python · cell 12

python

# 检查缺失值比例
train.isnull().sum().sort_values(ascending=False)

Output

Embarked       0
Cabin          0
Fare           0
Ticket         0
Parch          0
SibSp          0
Age            0
Sex            0
Name           0
Pclass         0
Survived       0
PassengerId    0
dtype: int64

In [ ]python · cell 13

python

任务三：编码分类变量

In [182]python · cell 15

python

# 取出所有的输入特征
data = train[['Pclass','Sex','Age','SibSp','Parch','Fare', 'Embarked']]

In [183]python · cell 16

python

# 进行虚拟变量转换
data = pd.get_dummies(data)

In [184]python · cell 17

python

data.head()

Output

   Pclass   Age  SibSp  Parch     Fare  Sex_female  Sex_male  Embarked_C  \
0       3  22.0      1      0   7.2500           0         1           0   
1       1  38.0      1      0  71.2833           1         0           1   
2       3  26.0      0      0   7.9250           1         0           0   
3       1  35.0      1      0  53.1000           1         0           0   
4       3  35.0      0      0   8.0500           0         1           0   

   Embarked_Q  Embarked_S  
0           0           1  
1           0           0  
2           0           1  
3           0           1  
4           0           1

.dataframe tbody tr th:only-of-type { vertical-align: middle; } .dataframe tbody tr th { vertical-align: top; } .dataframe thead th { text-align: right; }

	Pclass	Age	SibSp	Fare	Sex_female	Sex_male	Embarked_C	Embarked_S
0	3	22.0	1	7.2500	0	1	0	1
1	1	38.0	1	71.2833	1	0	1	0
2	3	26.0	0	7.9250	1	0	0	1
3	1	35.0	1	53.1000	1	0	0	1
4	3	35.0	0	8.0500	0	1	0	1

In [ ]python · cell 18

python

模型搭建

处理完前面的数据我们就得到建模数据，下一步是选择合适模型
在进行模型选择之前我们需要先知道数据集最终是进行监督学习还是无监督学习
除了根据我们任务来选择模型外，还可以根据数据样本量以及特征的稀疏性来决定
刚开始我们总是先尝试使用一个基本的模型来作为其baseline，进而再训练其他模型做对比，最终选择泛化能力或性能比较好的模型

思考0

数据集哪些差异会导致模型在拟合数据是发生变化

In [215]python · cell 22

python

# sklearn模型算法选择路径图
Image('20170624105439491.png')

Output

<IPython.core.display.Image object>

[省略较大 image/png 输出]

In [ ]python · cell 23

python

任务一：切割训练集和测试集

按比例切割训练集和测试集(一般测试集的比例有30%、25%、20%、15%和10%)
按目标变量分层进行等比切割
设置随机种子以便结果能复现

提示1

切割数据集是为了后续能评估模型泛化能力
sklearn中切割数据集的方法为train_test_split
查看函数文档可以在jupyter noteboo里面使用train_test_split?后回车即可看到
分层和随机种子在参数里寻找

思考1

什么情况下切割数据集的时候不用进行随机选取

In [216]python · cell 27

python

from sklearn.model_selection import train_test_split

In [217]python · cell 28

python

# 一般先取出X和y后再切割，有些情况会使用到未切割的，这时候X和y就可以用
X = data
y = train['Survived']

In [218]python · cell 29

python

# 对数据集进行切割
X_train, X_test, y_train, y_test = train_test_split(X, y, stratify=y, random_state=0)

In [219]python · cell 30

python

# 查看数据形状
X_train.shape, X_test.shape

Output

((668, 10), (223, 10))

In [ ]python · cell 31

python

任务二：模型创建

创建基于线性模型的分类模型（逻辑回归）
创建基于树的分类模型（决策树、随机森林）
查看模型的参数，并更改参数值，观察模型变化

提示2

逻辑回归不是回归模型而是分类模型，不要与LinearRegression混淆
随机森林其实是决策树集成为了降低决策树过拟合的情况
线性模型所在的模块为sklearn.linear_model
树模型所在的模块为sklearn.ensemble

思考2

为什么线性模型可以进行分类任务，背后是怎么的数学关系
对于多分类问题，线性模型是怎么进行分类的

In [223]python · cell 35

python

from sklearn.linear_model import LogisticRegression
from sklearn.ensemble import RandomForestClassifier

In [224]python · cell 36

python

# 默认参数逻辑回归模型
lr = LogisticRegression()
lr.fit(X_train, y_train)

Output

LogisticRegression(C=1.0, class_weight=None, dual=False, fit_intercept=True,
          intercept_scaling=1, max_iter=100, multi_class='ovr', n_jobs=1,
          penalty='l2', random_state=None, solver='liblinear', tol=0.0001,
          verbose=0, warm_start=False)

In [225]python · cell 37

python

# 查看训练集和测试集score值
print("Training set score: {:.2f}".format(lr.score(X_train, y_train)))
print("Testing set score: {:.2f}".format(lr.score(X_test, y_test)))

Output

Training set score: 0.80
Testing set score: 0.78

In [226]python · cell 38

python

# 调整参数后的逻辑回归模型
lr2 = LogisticRegression(C=100)
lr2.fit(X_train, y_train)

Output

LogisticRegression(C=100, class_weight=None, dual=False, fit_intercept=True,
          intercept_scaling=1, max_iter=100, multi_class='ovr', n_jobs=1,
          penalty='l2', random_state=None, solver='liblinear', tol=0.0001,
          verbose=0, warm_start=False)

In [227]python · cell 39

python

print("Training set score: {:.2f}".format(lr2.score(X_train, y_train)))
print("Testing set score: {:.2f}".format(lr2.score(X_test, y_test)))

Output

Training set score: 0.80
Testing set score: 0.79

In [228]python · cell 40

python

# 默认参数的随机森林分类模型
rfc = RandomForestClassifier()
rfc.fit(X_train, y_train)

Output

RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',
            max_depth=None, max_features='auto', max_leaf_nodes=None,
            min_impurity_decrease=0.0, min_impurity_split=None,
            min_samples_leaf=1, min_samples_split=2,
            min_weight_fraction_leaf=0.0, n_estimators=10, n_jobs=1,
            oob_score=False, random_state=None, verbose=0,
            warm_start=False)

In [229]python · cell 41

python

print("Training set score: {:.2f}".format(rfc.score(X_train, y_train)))
print("Testing set score: {:.2f}".format(rfc.score(X_test, y_test)))

Output

Training set score: 0.97
Testing set score: 0.82

In [230]python · cell 42

python

# 调整参数后的随机森林分类模型
rfc2 = RandomForestClassifier(n_estimators=100, max_depth=5)
rfc2.fit(X_train, y_train)

Output

RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',
            max_depth=5, max_features='auto', max_leaf_nodes=None,
            min_impurity_decrease=0.0, min_impurity_split=None,
            min_samples_leaf=1, min_samples_split=2,
            min_weight_fraction_leaf=0.0, n_estimators=100, n_jobs=1,
            oob_score=False, random_state=None, verbose=0,
            warm_start=False)

In [231]python · cell 43

python

print("Training set score: {:.2f}".format(rfc2.score(X_train, y_train)))
print("Testing set score: {:.2f}".format(rfc2.score(X_test, y_test)))

Output

Training set score: 0.86
Testing set score: 0.83

任务三：输出模型预测结果

输出模型预测分类标签
输出不通分类标签的预测概率

提示3

一般监督模型在sklearn里面有个predict能输出预测标签，predict_proba则可以输出标签概率

思考3

预测标签的概率对我们有什么帮助

In [234]python · cell 47

python

# 预测标签
pred = lr.predict(X_train)

In [235]python · cell 48

python

# 此时我们可以看到0和1的数组
pred[:10]

Output

array([0, 1, 1, 1, 0, 0, 1, 0, 1, 1], dtype=int64)

In [236]python · cell 49

python

# 预测标签概率
pred_proba = lr.predict_proba(X_train)

In [237]python · cell 50

python

pred_proba[:10]

Output

array([[0.62887291, 0.37112709],
       [0.14897206, 0.85102794],
       [0.47162003, 0.52837997],
       [0.20365672, 0.79634328],
       [0.86428125, 0.13571875],
       [0.9033887 , 0.0966113 ],
       [0.13829338, 0.86170662],
       [0.89516141, 0.10483859],
       [0.05735141, 0.94264859],
       [0.13593291, 0.86406709]])

In [ ]python · cell 51

python

模型评估

模型评估是为了知道模型的泛化能力。
交叉验证（cross-validation）是一种评估泛化性能的统计学方法，它比单次划分训练集和测试集的方法更加稳定、全面。
在交叉验证中，数据被多次划分，并且需要训练多个模型。
最常用的交叉验证是 k 折交叉验证（k-fold cross-validation），其中 k 是由用户指定的数字，通常取 5 或 10。
准确率（precision）度量的是被预测为正例的样本中有多少是真正的正例
召回率（recall）度量的是正类样本中有多少被预测为正类
f-分数是准确率与召回率的调和平均

任务一：交叉验证

用10折交叉验证来评估逻辑回归模型
计算交叉验证精度的平均值

In [245]python · cell 55

python

Image('Snipaste_2020-01-05_16-37-56.png')

Output

<IPython.core.display.Image object>

提示4

交叉验证在sklearn中的模块为sklearn.model_selection

思考4

k折越多的情况下会带来什么样的影响？

In [238]python · cell 58

python

from sklearn.model_selection import cross_val_score

In [200]python · cell 59

python

lr = LogisticRegression(C=100)
scores = cross_val_score(lr, X_train, y_train, cv=10)

In [201]python · cell 60

python

# k折交叉验证分数
scores

Output

array([0.82352941, 0.79411765, 0.80597015, 0.80597015, 0.8358209 ,
       0.88059701, 0.72727273, 0.86363636, 0.75757576, 0.71212121])

In [202]python · cell 61

python

# 平均交叉验证分数
print("Average cross-validation score: {:.2f}".format(scores.mean()))

Output

Average cross-validation score: 0.80

任务二：混淆矩阵

计算二分类问题的混淆矩阵
计算精确率、召回率以及f-分数

In [246]python · cell 63

python

Image('Snipaste_2020-01-05_16-38-26.png')

Output

<IPython.core.display.Image object>

In [247]python · cell 64

python

Image('Snipaste_2020-01-05_16-39-27.png')

Output

<IPython.core.display.Image object>

提示5

混淆矩阵的方法在sklearn中的sklearn.metrics模块
混淆矩阵需要输入真实标签和预测标签

思考5

如果自己实现混淆矩阵的时候该注意什么问题

In [147]python · cell 67

python

from sklearn.metrics import confusion_matrix

In [207]python · cell 68

python

# 训练模型
lr = LogisticRegression(C=100)
lr.fit(X_train, y_train)

Output

LogisticRegression(C=100, class_weight=None, dual=False, fit_intercept=True,
          intercept_scaling=1, max_iter=100, multi_class='ovr', n_jobs=1,
          penalty='l2', random_state=None, solver='liblinear', tol=0.0001,
          verbose=0, warm_start=False)

In [208]python · cell 69

python

# 模型预测结果
pred = lr.predict(X_train)

In [209]python · cell 70

python

# 混淆矩阵
confusion_matrix(y_train, pred)

Output

array([[350,  62],
       [ 71, 185]], dtype=int64)

In [210]python · cell 71

python

from sklearn.metrics import classification_report

In [211]python · cell 72

python

# 精确率、召回率以及f1-score
print(classification_report(y_train, pred))

Output

             precision    recall  f1-score   support

          0       0.83      0.85      0.84       412
          1       0.75      0.72      0.74       256

avg / total       0.80      0.80      0.80       668

In [ ]python · cell 73

python

任务三：ROC曲线

绘制ROC曲线

提示6

ROC曲线在sklearn中的模块为sklearn.metrics
ROC曲线下面所包围的面积越大越好

思考6

对于多分类问题如何绘制ROC曲线

In [212]python · cell 77

python

from sklearn.metrics import roc_curve

In [213]python · cell 78

python

fpr, tpr, thresholds = roc_curve(y_test, lr.decision_function(X_test))
plt.plot(fpr, tpr, label="ROC Curve")
plt.xlabel("FPR")
plt.ylabel("TPR (recall)")
# 找到最接近于0的阈值
close_zero = np.argmin(np.abs(thresholds))
plt.plot(fpr[close_zero], tpr[close_zero], 'o', markersize=10, label="threshold zero", fillstyle="none", c='k', mew=2)
plt.legend(loc=4)

Output

<matplotlib.legend.Legend at 0x2e4ea25db00>

<Figure size 720x432 with 1 Axes>

In [ ]python · cell 79

python

plt.rcParams['font.sans-serif'] = ['SimHei'] # 用来正常显示中文标签 plt.rcParams['axes.unicode_minus'] = False # 用来正常显示负号 plt.rcParams['figure.figsize'] = (10, 6) # 设置输出图片大小

PassengerId Survived Pclass \ 0 1 0 3 1 2 1 1 2 3 1 3 3 4 1 1 4 5 0 3 Name Sex Age SibSp \ 0 Braund, Mr. Owen Harris male 22.0 1 1 Cumings, Mrs. John Bradley (Florence Briggs Th... female 38.0 1 2 Heikkinen, Miss. Laina female 26.0 0 3 Futrelle, Mrs. Jacques Heath (Lily May Peel) female 35.0 1 4 Allen, Mr. William Henry male 35.0 0 Parch Ticket Fare Cabin Embarked 0 0 A/5 21171 7.2500 NaN S 1 0 PC 17599 71.2833 C85 C 2 0 STON/O2. 3101282 7.9250 NaN S 3 0 113803 53.1000 C123 S 4 0 373450 8.0500 NaN S

PassengerId

Survived

Pclass

Name

Sex

Age

SibSp

Parch

Ticket

Fare

Cabin

Embarked

Braund, Mr. Owen Harris

male

22.0

A/5 21171

7.2500

NaN

Cumings, Mrs. John Bradley (Florence Briggs Th...

female

38.0

PC 17599

71.2833

C85

Heikkinen, Miss. Laina

female

26.0

STON/O2. 3101282

7.9250

NaN

Futrelle, Mrs. Jacques Heath (Lily May Peel)

female

35.0

113803

53.1000

C123

Allen, Mr. William Henry

male

35.0

373450

8.0500

NaN

Embarked 0 Cabin 0 Fare 0 Ticket 0 Parch 0 SibSp 0 Age 0 Sex 0 Name 0 Pclass 0 Survived 0 PassengerId 0 dtype: int64

Pclass Age SibSp Parch Fare Sex_female Sex_male Embarked_C \ 0 3 22.0 1 0 7.2500 0 1 0 1 1 38.0 1 0 71.2833 1 0 1 2 3 26.0 0 0 7.9250 1 0 0 3 1 35.0 1 0 53.1000 1 0 0 4 3 35.0 0 0 8.0500 0 1 0 Embarked_Q Embarked_S 0 0 1 1 0 0 2 0 1 3 0 1 4 0 1

Pclass

Age

SibSp

Parch

Fare

Sex_female

Sex_male

Embarked_C

Embarked_Q

Embarked_S

22.0

7.2500

38.0

71.2833

26.0

7.9250

35.0

53.1000

35.0

8.0500

LogisticRegression(C=1.0, class_weight=None, dual=False, fit_intercept=True, intercept_scaling=1, max_iter=100, multi_class='ovr', n_jobs=1, penalty='l2', random_state=None, solver='liblinear', tol=0.0001, verbose=0, warm_start=False)

LogisticRegression(C=100, class_weight=None, dual=False, fit_intercept=True, intercept_scaling=1, max_iter=100, multi_class='ovr', n_jobs=1, penalty='l2', random_state=None, solver='liblinear', tol=0.0001, verbose=0, warm_start=False)

RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini', max_depth=None, max_features='auto', max_leaf_nodes=None, min_impurity_decrease=0.0, min_impurity_split=None, min_samples_leaf=1, min_samples_split=2, min_weight_fraction_leaf=0.0, n_estimators=10, n_jobs=1, oob_score=False, random_state=None, verbose=0, warm_start=False)

RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini', max_depth=5, max_features='auto', max_leaf_nodes=None, min_impurity_decrease=0.0, min_impurity_split=None, min_samples_leaf=1, min_samples_split=2, min_weight_fraction_leaf=0.0, n_estimators=100, n_jobs=1, oob_score=False, random_state=None, verbose=0, warm_start=False)

array([[0.62887291, 0.37112709], [0.14897206, 0.85102794], [0.47162003, 0.52837997], [0.20365672, 0.79634328], [0.86428125, 0.13571875], [0.9033887 , 0.0966113 ], [0.13829338, 0.86170662], [0.89516141, 0.10483859], [0.05735141, 0.94264859], [0.13593291, 0.86406709]])

precision recall f1-score support 0 0.83 0.85 0.84 412 1 0.75 0.72 0.74 256 avg / total 0.80 0.80 0.80 668

fpr, tpr, thresholds = roc_curve(y_test, lr.decision_function(X_test)) plt.plot(fpr, tpr, label="ROC Curve") plt.xlabel("FPR") plt.ylabel("TPR (recall)") # 找到最接近于0的阈值 close_zero = np.argmin(np.abs(thresholds)) plt.plot(fpr[close_zero], tpr[close_zero], 'o', markersize=10, label="threshold zero", fillstyle="none", c='k', mew=2) plt.legend(loc=4)