持续创作，加速生长！这是我参加「日新方案 10 月更文应战」的第24天，点击检查活动详情
文章换行有问题，望体谅
视频作者：菜菜TsaiTsai 链接：【技能干货】菜菜的机器学习sklearn【全85集】Python进阶_哔哩哔哩_bilibili
import pandas as pd
from sklearn.tree import DecisionTreeClassifier
import matplotlib.pyplot as plt
from sklearn.model_selection import GridSearchCV
from sklearn.model_selection import train_test_split
from sklearn.model_selection import cross_val_score
导入数据
data = pd.read_csv(r'D:\ObsidianWorkSpace\SklearnData\data.csv')
# 前面加r防止\转义，或者改\为/
探究数据
data.head(5) # 探究数据常用办法，head()里面是几就显示前几行，默以为5
---
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
data.info() # 探究数据常用办法
---
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 891 entries, 0 to 890
Data columns (total 12 columns):
PassengerId 891 non-null int64
Survived 891 non-null int64
Pclass 891 non-null int64
Name 891 non-null object
Sex 891 non-null object
Age 714 non-null float64
SibSp 891 non-null int64
Parch 891 non-null int64
Ticket 891 non-null object
Fare 891 non-null float64
Cabin 204 non-null object
Embarked 889 non-null object
dtypes: float64(2), int64(5), object(5)
memory usage: 83.6+ KB
挑选特征。删去缺失过多的列，自己判断以为与预测成果没有关系的列 Name和Tickle和是否存活无关，Cabin缺失过多
data.drop(['Cabin','Name','Ticket'],axis=1,inplace=True)# 用drop删去数据
# 默认axis=0，沿行的方向删去，axis=1，沿列的方向删去
# 列表中写要删去的列名/行名
# inplace:是否用删去后的表掩盖原表，True替换，False不替换。用data = data.drop(inplace=False)也可以替换
处理缺失值。对缺失值较多的列进行添补，有一些特征只缺失一两个值，可以采取直接删去记载的办法
data['Age'] = data['Age'].fillna(data['Age'].mean())
data = data.dropna() # 什么也不填，即只需有缺失值就整行删去
dropna(axis=0, how='any', thresh=None, subset=None, inplace=False)
# axis:轴向，默认0删去行，1删去列
# inplace:是否掩盖原表，假如不掩盖，则回来新表，假如掩盖，无回来值
# how:默认'any'即所内行/列有一个空值就删去，'all'即所内行/列全为空值才删去
# thresh:填int，当非空值数量大于等于该值就保存
# subset:填列表，当axis=0，列表元素为列标签，对每一行数据，对应列假如有空值，则删去；当axis=1，列表元素为行标签，对每一列数据，对应行假如有空值，则删去
df = pd.DataFrame([[1,2,3,None],[None,5,6,7]])
print(df)
df.dropna(axis=1, subset=[0])
---
0 1 2 3
0 1.0 2 3 NaN
1 NaN 5 6 7.0
0 1 2
0 1.0 2 3
1 NaN 5 6
data.info()
---
<class 'pandas.core.frame.DataFrame'>
Int64Index: 889 entries, 0 to 890
Data columns (total 9 columns):
PassengerId 889 non-null int64
Survived 889 non-null int64
Pclass 889 non-null int64
Sex 889 non-null object
Age 889 non-null float64
SibSp 889 non-null int64
Parch 889 non-null int64
Fare 889 non-null float64
Embarked 889 non-null object
dtypes: float64(2), int64(5), object(2)
memory usage: 69.5+ KB
咱们需求将分类变量转换为数值型变量。也便是object转换为num Sex和Embarked是Object，需求改为数字
labels = data['Embarked'].unique().tolist()
data['Embarked'] = data['Embarked'].apply(lambda x:labels.index(x))
# 要求该特征不同取值无关联，一般unique以后数量少于10个，可以这样处理
data['Embarked'].unique() # unique取出所有的值并删去重复值
---
array(['S', 'C', 'Q'], dtype=object)
data['Sex'] = (data['Sex'] == 'male').astype('int')
# astype能够将一个pandas对象转换为某种类型
# 和apply(int(x))不同，astype可以将文本类转换为数字，用这个方法可以很快捷的将二分类特征转换为0-1
这里也可以用
data.loc[:,'Sex']data.loc[:,'Sex']根据标签的索引，后边只能加列名或列名切片，不接收数字（列号）data.iloc[:,3]即根据位置的索引，后边只能加数字或数字切片 二者都可以接收布尔索引
提取标签和特征矩阵，分离测试集和练习集
x = data.iloc[:,data.columns != 'Survived']
y = data.iloc[:,data.columns == 'Survived']
Xtrain, Xtest, Ytrain, Ytest = train_test_split(x, y, test_size=0.3)
Xtrain.head()
---
PassengerId Pclass Sex Age SibSp Parch Fare Embarked
846 847 3 1 29.699118 8 2 69.5500 0
858 859 3 0 24.000000 0 3 19.2583 1
564 565 3 0 29.699118 0 0 8.0500 0
220 221 3 1 16.000000 0 0 8.0500 0
774 775 2 0 54.000000 1 3 23.0000 0
修正索引
假如索引乱了，而且咱们不是有意让它乱，最好恢复成0-shape[0]的方式
for i in [Xtrain, Xtest, Ytrain, Ytest]:
i.index = range(i.shape[0])
导入模型，大略跑一下检查成果
clf = DecisionTreeClassifier(random_state=0)
clf = clf.fit(Xtrain,Ytrain)
score = clf.score(Xtest,Ytest)
score
---
0.7265917602996255
clf = DecisionTreeClassifier(random_state=0)
score = cross_val_score(clf,x,y,cv=10).mean()
score
---
0.7571118488253319
在不同max_depth下观察模型的拟合情况
tr = []
te = []
for i in range(10):
clf = DecisionTreeClassifier(random_state=0
,max_depth=i+1
)
clf = clf.fit(Xtrain,Ytrain)
score_tr = clf.score(Xtrain,Ytrain)
score_te = cross_val_score(clf,x,y,cv=10).mean()
tr.append(score_tr)
te.append(score_te)
print(max(te))
plt.figure()
plt.plot(range(1,11),tr,color='red',label='train')
plt.plot(range(1,11),te,color='blue',label='test')
plt.xticks(range(1,11))
plt.legend()
plt.show()
咱们考虑用entropy。之前说entropy简单过拟合，但这是对练习集来说的，对测试集不一定，且此刻练习集拟合程度也不太好
tr = []
te = []
for i in range(10):
clf = DecisionTreeClassifier(random_state=0
,max_depth=i+1
,criterion='entropy'
)
clf = clf.fit(Xtrain,Ytrain)
score_tr = clf.score(Xtrain,Ytrain)
score_te = cross_val_score(clf,x,y,cv=10).mean()
tr.append(score_tr)
te.append(score_te)
print(max(te))
plt.figure()
plt.plot(range(1,11),tr,color='red',label='train')
plt.plot(range(1,11),te,color='blue',label='test')
plt.xticks(range(1,11))
plt.legend()
plt.show()
网格查找：能够帮助咱们同时调整多个参数的技能，是一种枚举技能 给定字典，字典中有参数规模，找到参数规模内能让模型最好的效果的组合 由于是多个参数组合，看起来像网格，所以叫网格查找；由于是枚举，所以计算量很大，要注意参数数量，界定好参数规模 实际上，大都时刻参数挑选什么都是依靠自己的判断，或者只跑一两个参数组合
# 一串参数和这些参数对应的咱们期望网格查找来查找的参数取值规模
gini_thresholds = np.linspace(0,0.5,5)
# 原课程用的50，我就不折腾我这破铜烂铁了
parameters = {'criterion':('gini','entropy')
,'splitter':('best','random')
,'max_depth':[*range(2,6)]
,'min_samples_leaf':[*range(5,50,5)]
,'min_impurity_decrease':[*gini_thresholds]
}
clf = DecisionTreeClassifier(random_state=0)
GS = GridSearchCV(clf, parameters, cv=10)# 同时有fit，score，交叉验证三种功用
GS = GS.fit(Xtrain,Ytrain)
GS.best_params_ # 从咱们输入的参数和参数取值的列表中，回来最佳组合
---
{'criterion': 'gini',
'max_depth': 6,
'min_impurity_decrease': 0.0,
'min_samples_leaf': 11,
'splitter': 'best'}
GS.best_score_ # 网格查找后的模型评判规范
---
0.819935691318328
现在挑选参数的办法有：学习曲线和网格查找
