Kaggle泰坦尼克号预测——Last

整理了一下全部流程，写成了一个ipynb文件

导入数据集

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
import warnings
warnings.filterwarnings('ignore')

##载入数据集
data_path = 'D:/Now/Titanic/'

train_data = pd.read_csv(data_path + 'train.csv',sep = ',')
test_data = pd.read_csv(data_path + 'test.csv',sep = ',')

train_data.info()

数据探索性分析

##查看缺失值
train_miss_data = train_data.isnull().sum()

train_miss_data = train_miss_data[train_miss_data>0]
train_miss_data.plot.bar()

在train_data的缺失值中，可以看到Cabin的缺失值比较多。Age的缺失值相对来说也比较多

train_data.head(10)

Pclass,Cabin,Embarked,Ticket,Sex,Survived是定性数据，Fare,SibSp,Parch,Age是定量数据。

查看Embarked数据
查阅资料得知：泰坦尼克号从英国南安普敦(Southampton)港出发，途经法国瑟堡-奥克特维尔(Cherbourg-Octeville)以及爱尔兰昆士敦(Queenstown)，计划中的目的地为美国纽约(New York)，开始了这艘“梦幻客轮”的处女航。

train_data['Embarked'].value_counts()

将Embarked转换成数值型数据

#train_data将S转为1，C转为2，Q转为3
train_data['Embarked'].replace('S',1,inplace=True)
train_data['Embarked'].replace('C',2,inplace=True)
train_data['Embarked'].replace('Q',3,inplace=True)
train_data['Embarked'].value_counts()

'''
1.0    644
2.0    168
3.0     77
Name: Embarked, dtype: int64
'''

#test_data将S转为1，C转为2，Q转为3
test_data['Embarked'].replace('S',1,inplace=True)
test_data['Embarked'].replace('C',2,inplace=True)
test_data['Embarked'].replace('Q',3,inplace=True)
test_data['Embarked'].value_counts()
'''
1    270
2    102
3     46
Name: Embarked, dtype: int64
'''

查看Sex数据

train_data['Sex'].value_counts()

将Sex转换成数值型数据

#train_data将male转换成1，female转换成2
train_data['Sex'].replace('male',1,inplace=True)
train_data['Sex'].replace('female',2,inplace=True)
train_data['Sex'].value_counts()
'''
1    577
2    314
Name: Sex, dtype: int64
'''

#test_data将male转换成1，female转换成2
test_data['Sex'].replace('male',1,inplace=True)
test_data['Sex'].replace('female',2,inplace=True)
test_data['Sex'].value_counts()
'''
1    266
2    152
Name: Sex, dtype: int64
'''

转换之后查看相关性

abs(train_data.corr(method='pearson')).plot(kind='bar')
plt.legend(loc = 'best')
abs(train_data.corr(method='pearson'))

查看存活情况和其他数据的相关性

abs(train_data.corr(method='pearson'))['Survived']

可以看到，Sex，Pclass，Fare与Survived都有较明显的相关性

换个图看看：

sns.heatmap(abs(train_data.corr(method='kendall')))

train_data['Survived'].value_counts(normalize=True)
#空准确率 有38%的人获救

##对每一列绘制图观察
cols = ['Age','SibSp','Parch','Fare']
for i in cols:
    sns.distplot(train_data[train_data['Survived']==0][i],label='notSurvived',kde=False)
    sns.distplot(train_data[train_data['Survived']==1][i],label='Survived',kde=False)
#     plt.hist(train_data[train_data['Survived']==0][i],10,alpha = 0.5,label='notSurvived')
#     plt.hist(train_data[train_data['Survived']==1][i],10,alpha = 0.5,label='Survived')
    plt.legend(loc = 'best')
    plt.xlabel(i)
    plt