Author： 马肖
E-Mail： maxiaoscut@aliyun.com
GitHub： https://github.com/Albertsr

In [1]:

import warnings
warnings.filterwarnings("ignore")

import numpy as np
import pandas as pd
import xgboost as xgb
import seaborn as sns
import matplotlib.pyplot as plt


sns.set_style('whitegrid')
sns.set(font_scale=1.2, style='ticks', palette='Set2') 
%matplotlib inline

In [2]:

train = pd.read_csv("DEA.csv")
train.head(3)

Out[2]:

	PassengerId	Survived	Pclass	Sex	Age	SibSp	Fare	Embarked
0	1	0	3	male	22.0	1	7.2500	S
1	2	1	1	female	38.0	1	71.2833	C
2	3	1	3	female	26.0	0	7.9250	S

1 对离散型特征的探索（以Pclass为例）¶

1.1 各等级舱位等级的人数¶

In [3]:

train["Pclass"].value_counts()

Out[3]:

3    491
1    216
2    184
Name: Pclass, dtype: int64

In [4]:

sns.countplot(x='Pclass', data=train)
sns.despine()

结论：三等舱人人数最多，二等舱人数最少¶

1.2 船舱等级对生还率的影响¶

In [5]:

train.pivot_table(index=["Pclass"], values=["Survived"], aggfunc="mean")
# 等价于train.groupby(["Embarked"])["Survived"].mean()

Out[5]:

	Survived
Pclass
1	0.629630
2	0.472826
3	0.242363

In [6]:

sns.barplot(x='Pclass', y='Survived', data=train)
sns.despine()

结论：一等舱乘客有最高的生还率¶

1.3 与其他离散型变量的联合探索¶

1.3.1 对Pclass、Sex联合探索：无论哪个船舱等级，女性都有更高的生还率¶

In [7]:

train.pivot_table(index=["Pclass"], values=["Survived"], columns=["Sex"])

Out[7]:

	Survived
Sex	female	male
Pclass
1	0.968085	0.368852
2	0.921053	0.157407
3	0.500000	0.135447

In [8]:

sns.barplot(x='Pclass', y='Survived', data=train, hue='Sex', dodge=True)  # hue_order=['female','male']
sns.despine()
plt.legend()

Out[8]:

<matplotlib.legend.Legend at 0x214388814a8>

1.3.2 不同性别对应的不同舱位均值 : 男性乘客的平均舱位稍高于女性乘客¶

In [9]:

train.pivot_table(index=["Sex"], values=["Pclass"])

Out[9]:

	Pclass
Sex
female	2.159236
male	2.389948

In [10]:

sns.barplot(x='Sex', y='Pclass', data=train) 
sns.despine()

1.3.3 对Pclass、Embarked进行联合探索：S港口登录的乘客生还率最低，C港口登录的乘客生还率稍高于Q¶

In [11]:

train.pivot_table(index=["Pclass"], values=["Survived"], columns=["Embarked"])

Out[11]:

	Survived
Embarked	C	Q	S
Pclass
1	0.694118	0.500000	0.589147
2	0.529412	0.666667	0.463415
3	0.378788	0.375000	0.189802

In [12]:

sns.barplot(x='Pclass', y='Survived', data=train, hue='Embarked', dodge=True)  # hue_order=['female','male']
sns.despine()
plt.legend()

Out[12]:

<matplotlib.legend.Legend at 0x2143895b908>

1.4 与其他连续性变量的联合探索¶

结论1：无论什么性别，舱位与年龄成反比；无论哪个舱位，男性要年长于女性¶

In [13]:

train.pivot_table(index=["Pclass"], values=["Age", 'Fare'], columns=["Sex"])

Out[13]:

	Age		Fare
Sex	female	male	female	male
Pclass
1	34.768107	41.711241	106.125798	67.226127
2	28.744443	30.955625	21.970121	19.741782
3	21.973566	26.707903	16.118810	12.661633

In [14]:

sns.barplot(x='Pclass', y='Age', data=train, hue="Sex", dodge=True, errwidth=2)  # hue_order=['female','male']
sns.despine()
plt.legend(loc=1)

Out[14]:

<matplotlib.legend.Legend at 0x214389c3c18>

结论2：舱位与票价成正比；无论哪个舱位，女性支付的费用都高于男性¶

In [15]:

train.pivot_table(index=["Pclass"], values=["Fare"], columns=["Sex"])

Out[15]:

	Fare
Sex	female	male
Pclass
1	106.125798	67.226127
2	21.970121	19.741782
3	16.118810	12.661633

In [16]:

sns.barplot(x='Pclass', y='Fare', data=train, hue="Sex", dodge=True, errwidth=2)  # hue_order=['female','male']
sns.despine()
plt.legend(loc=1)

Out[16]:

<matplotlib.legend.Legend at 0x21438a52d30>

将图表予以整合¶

In [17]:

sns.set(font_scale=1.2,style='ticks',palette='Set2') 
f, axes = plt.subplots(2, 2, figsize=(10, 8))
sns.despine()


#特征的不同取值对应的样本数量
sns.countplot(x='Embarked', data=train, ax=axes[0, 0])

#特征Embarked的不同取值对应于不同的类标的数量
sns.countplot(x='Survived', hue="Embarked", data=train, order=[1,0], ax=axes[0, 1])

#特征的不同取值对应的生还概率
sns.barplot(x='Embarked', y='Survived',data=train, dodge=True, ax=axes[1, 0])    #order=['S','C','Q']


# 联合其他特征Sex共同取各种类别的概率
# dodge=True表示不同性别予以分开展示，否则为堆叠展示
sns.barplot(x='Embarked', y='Survived',data=train, hue='Sex',hue_order=['female','male'], dodge=True, ax=axes[1, 1]) 
plt.legend(loc='upper right')

Out[17]:

<matplotlib.legend.Legend at 0x21438a85550>

2 对连续型特征进行探索¶

2.1 直方图、KDE曲线、箱型图¶

通过箱型图发现Fare特征存在一些极端的异常值，对这些特征修改为95分位点的值¶

In [18]:

f, axes = plt.subplots(1, 2, figsize=(10, 4))
sns.boxplot(train["Fare"], ax=axes[0])
sns.stripplot(train["Fare"], jitter=True, ax=axes[0])
axes[0].set_title("Fare_Original", fontsize=13)
axes[0].set_xlabel('')

Fare_per_95 = np.percentile(train["Fare"], 95)
train["Fare"][train["Fare"] >= Fare_per_95] = Fare_per_95
sns.boxplot(train["Fare"], ax=axes[1])
sns.stripplot(train["Fare"], jitter=True, ax=axes[1])
axes[1].set_title("Fare_processed", fontsize=13)
axes[1].set_xlabel('')
plt.show()

In [19]:

sns.set(font_scale=1.2,style='ticks',palette='Set2') 
f, axes = plt.subplots(2, 2, figsize=(10, 8))
sns.despine()


#特征的不同取值对应的样本数量
sns.distplot(train['Age'], ax=axes[0,0])
axes[0,0].set_title("Age_KDE", fontsize=12)
axes[0,0].set_xlabel("")

sns.boxplot(train['Age'], ax=axes[0,1])
axes[0,1].set_title("Age_Boxplot", fontsize=12)
axes[0,1].set_xlabel("")

#特征Embarked的不同取值对应于不同的类标的数量
sns.distplot(train['Fare'], ax=axes[1,0])
axes[1,0].set_title("Fare_KDE", fontsize=12)
axes[1,0].set_xlabel("")

sns.boxplot(train['Fare'], ax=axes[1,1])
axes[1,1].set_title("Fare_Boxplot", fontsize=12)
axes[1,1].set_xlabel("")

Out[19]:

Text(0.5,0,'')

In [20]:

fig = plt.figure(figsize=(12,8))

sns.distplot(train['Age'], ax=plt.subplot(221))
plt.title("Age_KDE", fontsize=12)
plt.xlabel('')


sns.boxplot(train['Age'], ax=plt.subplot(222))
plt.title("Age_Boxplot", fontsize=12)
plt.xlabel('')


sns.distplot(train['Fare'], ax=plt.subplot(223))
plt.title("Fare_KDE", fontsize=12)
plt.xlabel('')


sns.boxplot(train['Fare'], ax=plt.subplot(224))
plt.title("Fare_Boxplot", fontsize=12)
plt.xlabel('')

sns.despine()

2.2 不同分类结果对应的连续型变量的KDE曲线¶

In [21]:

f, axes = plt.subplots(1, 2, figsize=(10, 4))

sns.distplot(train['Age'][train['Survived']==0], kde_kws={"color": "k", "lw": 2, "label": "UnSurvived"}, ax=axes[0])
sns.distplot(train['Age'][train['Survived']==1], kde_kws={"color": "r", "lw": 2, "label": "Survived"}, ax=axes[0])

sns.distplot(train['Fare'][train['Survived']==0],kde_kws={"color": "k", "lw": 2, "label": "UnSurvived"}, ax=axes[1])
sns.distplot(train['Fare'][train['Survived']==1],kde_kws={"color": "r", "lw": 2, "label": "Survived"}, ax=axes[1])

sns.despine()

In [22]:

f, axes = plt.subplots(2, 2, figsize=(14, 8))

sns.distplot(train['Age'][(train['Survived']==0) & (train['Sex']=="male")], 
             kde_kws={"color": "k", "lw": 2, "label": "UnSurvived_male"}, ax=axes[0, 0])

sns.distplot(train['Age'][(train['Survived']==1) & (train['Sex']=="male")], 
             kde_kws={"color": "r", "lw": 2, "label": "Survived_male"}, ax=axes[0, 0])



sns.distplot(train['Age'][(train['Survived']==0) & (train['Sex']=="female")], 
             kde_kws={"color": "k", "lw": 2, "label": "UnSurvived_female"}, ax=axes[0, 1])

sns.distplot(train['Age'][(train['Survived']==1) & (train['Sex']=="female")], 
             kde_kws={"color": "r", "lw": 2, "label": "Survived_female"}, ax=axes[0, 1])

# Fare
sns.distplot(train['Fare'][(train['Survived']==0) & (train['Sex']=="male")], 
             kde_kws={"color": "k", "lw": 2, "label": "UnSurvived_male"}, ax=axes[1, 0])

sns.distplot(train['Fare'][(train['Survived']==1) & (train['Sex']=="male")], 
             kde_kws={"color": "r", "lw": 2, "label": "Survived_male"}, ax=axes[1, 0])


sns.distplot(train['Fare'][(train['Survived']==0) & (train['Sex']=="female")], 
             kde_kws={"color": "k", "lw": 2, "label": "UnSurvived_female"}, ax=axes[1, 1])

sns.distplot(train['Fare'][(train['Survived']==1) & (train['Sex']=="female")], 
             kde_kws={"color": "r", "lw": 2, "label": "Survived_female"}, ax=axes[1, 1])

sns.despine()

结论：¶

1.生还的人，无论什么性别，平均来看，支付了更高的费用，也比较年长

2.可以利用KDE曲线对Fare/Age做离散化处理

2.3 与其他的特征的联合探索¶

In [23]:

f, axes = plt.subplots(1, 2, figsize=(10, 3.5))
sns.barplot(x=train["Survived"], y=train["Fare"], ax=axes[0])
sns.barplot(x=train["Survived"], y=train["Age"], ax=axes[1])
sns.despine()

In [24]:

f, axes = plt.subplots(1, 2, figsize=(12, 4))
sns.barplot(x=train["Survived"], y=train["Fare"], hue=train["Sex"], ax=axes[0])
sns.barplot(x=train["Survived"], y=train["Age"], hue=train["Sex"], ax=axes[1])

plt.legend()
sns.despine()

3 SibSp Parch 这些特征虽然是连续型变量，但是取值较少，分布不均匀，可作为离散型变量处理¶

In [25]:

f, axes = plt.subplots(2, 3, figsize=(14, 8))
sns.countplot(x=train["Parch"], ax=axes[0,0])
sns.pointplot(x="Parch", y="Survived", data=train, ax=axes[0,1])
sns.barplot(x="Parch", y="Survived", data=train, ax=axes[0,2])

sns.countplot(x=train["SibSp"], ax=axes[1,0])
sns.pointplot(x="SibSp", y="Survived", data=train, ax=axes[1,1])
sns.barplot(x="SibSp", y="Survived", data=train, ax=axes[1,2])

sns.despine()