第十章 决策树与随机森林

# 导入第三方模块
import pandas as pd
# 读入数据
Titanic = pd.read_csv(r'Titanic.csv')
Titanic.head()

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

# 删除无意义的变量，并检查剩余自字是否含有缺失值
Titanic.drop(['PassengerId','Name','Ticket','Cabin'], axis = 1, inplace = True)
Titanic.isnull().sum(axis = 0)

Survived      0
Pclass        0
Sex           0
Age         177
SibSp         0
Parch         0
Fare          0
Embarked      2
dtype: int64

# 对Sex分组，用各组乘客的平均年龄填充各组中的缺失年龄
fillna_Titanic = []
for i in Titanic.Sex.unique():
    update = Titanic.loc[Titanic.Sex == i,].fillna(value = {'Age': Titanic.Age[Titanic.Sex == i].mean()}, inplace = False)
    fillna_Titanic.append(update)
Titanic = pd.concat(fillna_Titanic)
# 使用Embarked变量的众数填充缺失值
Titanic.fillna(value = {'Embarked':Titanic.Embarked.mode()[0]}, inplace=True)
Titanic.head()

	Survived	Pclass	Sex	Age	SibSp	Parch	Fare	Embarked
0	0	3	male	22.000000	1	0	7.2500	S
4	0	3	male	35.000000	0	0	8.0500	S
5	0	3	male	30.726645	0	0	8.4583	Q
6	0	1	male	54.000000	0	0	51.8625	S
7	0	3	male	2.000000	3	1	21.0750	S

# 将数值型的Pclass转换为类别型，否则无法对其哑变量处理
Titanic.Pclass = Titanic.Pclass.astype('category')
# 哑变量处理
dummy = pd.get_dummies(Titanic[['Sex','Embarked','Pclass']])
# 水平合并Titanic数据集和哑变量的数据集
Titanic = pd.concat([Titanic,dummy], axis = 1)
# 删除原始的Sex、Embarked和Pclass变量
Titanic.drop(['Sex','Embarked','Pclass'], inplace=True, axis = 1)
Titanic.head()

	Survived	Age	SibSp	Parch	Fare	Sex_female	Sex_male	Embarked_C	Embarked_Q	Embarked_S	Pclass_1	Pclass_2	Pclass_3
0	0	22.000000	1	0	7.2500	0	1	0	0	1	0	0	1
4	0	35.000000	0	0	8.0500	0	1	0	0	1	0	0	1
5	0	30.726645	0	0	8.4583	0	1	0	1	0	0	0	1
6	0	54.000000	0	0	51.8625	0	1	0	0	1	1	0	0
7	0	2.000000	3	1	21.0750	0	1	0	0	1	0	0	1

# 导入第三方包
from sklearn import model_selection
# 取出所有自变量名称
predictors = Titanic.columns[1:]
# 将数据集拆分为训练集和测试集，且测试集的比例为25%
X_train, X_test, y_train, y_test = model_selection.train_test_split(Titanic[predictors], Titanic.Survived, 
                                                                    test_size = 0.25, random_state = 1234)

# 导入第三方模块
from sklearn.model_selection import GridSearchCV
from sklearn import tree
# 预设各参数的不同选项值
max_depth = [2,3,4,5,6]
min_samples_split = [2,4,6,8]
min_samples_leaf = [2,4,8,10,12]
# 将各参数值以字典形式组织起来
parameters = {'max_depth':max_depth, 'min_samples_split':min_samples_split, 'min_samples_leaf':min_samples_leaf}
# 网格搜索法，测试不同的参数值
grid_dtcateg = GridSearchCV(estimator = tree.DecisionTreeClassifier(), param_grid = parameters, cv=10)
# 模型拟合
grid_dtcateg.fit(X_train, y_train)
# 返回最佳组合的参数值
grid_dtcateg.best_params_

{'max_depth': 3, 'min_samples_leaf': 4, 'min_samples_split': 2}

# 导入第三方模块
from sklearn import metrics
# 构建分类决策树
CART_Class = tree.DecisionTreeClassifier(max_depth=3, min_samples_leaf = 4, min_samples_split=2)
# 模型拟合
decision_tree = CART_Class.fit(X_train, y_train)
# 模型在测试集上的预测
pred = CART_Class.predict(X_test)
# 模型的准确率
print('模型在测试集的预测准确率：\n',metrics.accuracy_score(y_test, pred))

模型在测试集的预测准确率：
 0.8295964125560538

# 导入第三方包
import matplotlib.pyplot as plt
y_score = CART_Class.predict_proba(X_test)[:,1]
fpr,tpr,threshold = metrics.roc_curve(y_test, y_score)
# 计算AUC的值
roc_auc = metrics.auc(fpr,tpr)

# 绘制面积图
plt.stackplot(fpr, tpr, color='steelblue', alpha = 0.5, edgecolor = 'black')
# 添加边际线
plt.plot(fpr, tpr, color='black', lw = 1)
# 添加对角线
plt.plot([0,1],[0,1], color = 'red', linestyle = '--')
# 添加文本信息
plt.text(0.5,0.3,'ROC curve (area = %0.2f)' % roc_auc)
# 添加x轴与y轴标签
plt.xlabel('1-Specificity')
plt.ylabel('Sensitivity')
# 显示图形
plt.show()

<Figure size 640x480 with 1 Axes>

# 需要在电脑中安装Graphviz
# https://graphviz.gitlab.io/_pages/Download/Download_windows.html
# 然后将解压文件中的bin设置到环境变量中
# 导入第三方模块
from sklearn.tree import export_graphviz
from IPython.display import Image
import pydotplus
from sklearn.externals.six import StringIO
# 绘制决策树
dot_data = StringIO()
export_graphviz(
    decision_tree,
    out_file=dot_data,  
    feature_names=predictors,
    class_names=['Unsurvived','Survived'],  
    # filled=True,
    rounded=True,  
    special_characters=True
)
# 决策树展现
graph = pydotplus.graph_from_dot_data(dot_data.getvalue())
Image(graph.create_png()) 

---------------------------------------------------------------------------

InvocationException                       Traceback (most recent call last)

<ipython-input-10-35271d1803f9> in <module>()
     20 # 决策树展现
     21 graph = pydotplus.graph_from_dot_data(dot_data.getvalue())
---> 22 Image(graph.create_png())


~\Anaconda3\lib\site-packages\pydotplus\graphviz.py in <lambda>(f, prog)
   1795             self.__setattr__(
   1796                 'create_' + frmt,
-> 1797                 lambda f=frmt, prog=self.prog: self.create(format=f, prog=prog)
   1798             )
   1799             f = self.__dict__['create_' + frmt]


~\Anaconda3\lib\site-packages\pydotplus\graphviz.py in create(self, prog, format)
   1958             if self.progs is None:
   1959                 raise InvocationException(
-> 1960                     'GraphViz\'s executables not found')
   1961 
   1962         if prog not in self.progs:


InvocationException: GraphViz's executables not found

# 导入第三方包
from sklearn import ensemble
# 构建随机森林
RF_class = ensemble.RandomForestClassifier(n_estimators=200, random_state=1234)
# 随机森林的拟合
RF_class.fit(X_train, y_train)
# 模型在测试集上的预测
RFclass_pred = RF_class.predict(X_test)
# 模型的准确率
print('模型在测试集的预测准确率：\n',metrics.accuracy_score(y_test, RFclass_pred))

# 计算绘图数据
y_score = RF_class.predict_proba(X_test)[:,1]
fpr,tpr,threshold = metrics.roc_curve(y_test, y_score)
roc_auc = metrics.auc(fpr,tpr)
# 绘图
plt.stackplot(fpr, tpr, color='steelblue', alpha = 0.5, edgecolor = 'black')
plt.plot(fpr, tpr, color='black', lw = 1)
plt.plot([0,1],[0,1], color = 'red', linestyle = '--')
plt.text(0.5,0.3,'ROC curve (area = %0.2f)' % roc_auc)
plt.xlabel('1-Specificity')
plt.ylabel('Sensitivity')
plt.show()

# 变量的重要性程度值
importance = RF_class.feature_importances_
# 构建含序列用于绘图
Impt_Series = pd.Series(importance, index = X_train.columns)
# 对序列排序绘图
Impt_Series.sort_values(ascending = True).plot('barh')
plt.show()

# 读入数据
NHANES = pd.read_excel(r'C:\Users\Administrator\Desktop\NHANES.xlsx')
NHANES.head()
print(NHANES.shape)

# 取出自变量名称
predictors = NHANES.columns[:-1]
# 将数据集拆分为训练集和测试集
X_train, X_test, y_train, y_test = model_selection.train_test_split(NHANES[predictors], NHANES.CKD_epi_eGFR, 
                                                                    test_size = 0.25, random_state = 1234)

# 预设各参数的不同选项值
max_depth = [18,19,20,21,22]
min_samples_split = [2,4,6,8]
min_samples_leaf = [2,4,8]
parameters = {'max_depth':max_depth, 'min_samples_split':min_samples_split, 'min_samples_leaf':min_samples_leaf}
# 网格搜索法，测试不同的参数值
grid_dtreg = GridSearchCV(estimator = tree.DecisionTreeRegressor(), param_grid = parameters, cv=10)
# 模型拟合
grid_dtreg.fit(X_train, y_train)
# 返回最佳组合的参数值
grid_dtreg.best_params_

# 构建用于回归的决策树
CART_Reg = tree.DecisionTreeRegressor(max_depth = 20, min_samples_leaf = 2, min_samples_split = 4)
# 回归树拟合
CART_Reg.fit(X_train, y_train)
# 模型在测试集上的预测
pred = CART_Reg.predict(X_test)
# 计算衡量模型好坏的MSE值
metrics.mean_squared_error(y_test, pred)

# 构建用于回归的随机森林
RF = ensemble.RandomForestRegressor(n_estimators=200, random_state=1234)
# 随机森林拟合
RF.fit(X_train, y_train)
# 模型在测试集上的预测
RF_pred = RF.predict(X_test)
# 计算模型的MSE值
metrics.mean_squared_error(y_test, RF_pred)

# 构建变量重要性的序列
importance = pd.Series(RF.feature_importances_, index = X_train.columns)
# 排序并绘图
importance.sort_values().plot('barh')
plt.show()