타이타닉 프로젝트 (1) - 자료 분석

first, load the csv file from the kaggle

In[1]:

import pandas as pd

​

_train = pd.read_csv('../input/titanic/train.csv')

_test = pd.read_csv('../input/titanic/test.csv')

​

train = _train.copy()

test = _test.copy()

to find out if there is NaN values in the Dataframe, I use df.isnull().sum() function

In[2]:

train.isnull().sum()

Out[2]:

PassengerId      0
Survived         0
Pclass           0
Name             0
Sex              0
Age            177
SibSp            0
Parch            0
Ticket           0
Fare             0
Cabin          687
Embarked         2
dtype: int64

In case of "Cabin" column, there are too many NaN values, which could make prediction improper

Thus, I erased the column from the dataframe

In[3]:

train.drop('Cabin', axis=1, inplace=True)

Moreover, I have to modify 'Age' column and 'Embarked' column using some values. Let's think about it later.

Since I wanted to find out any relationship between some features and Survival Rate I made a function which displays survival rate per some features

In[4]:

import matplotlib.pyplot as plt

import seaborn as sns

​

​

def bar_survival(feature):

    _survived = train[train['Survived']==1][feature].value_counts()

    _death = train[train['Survived']==0][feature].value_counts()

​

    survived = _survived.rename("Survived")

    death = _death.rename("Death")

    data = pd.DataFrame([survived, death]).transpose()

    Percentage = []    

    for i in range(data.shape[0]):

        tmp_sum = data.iloc[i].sum()        

        tmp_sur = data.iloc[i][0]        

        tmp_per = round(tmp_sur/tmp_sum*100, 0)

        Percentage.append(tmp_per)        

    data.loc[: , feature] = pd.Series(Percentage, index=data.index)

    data_per = data.drop(['Survived', 'Death'],axis=1)

    data_per.plot(kind='bar', stacked=True, figsize=(5,5))

In[5]:

bar_survival('Pclass')

it seems that 'Pclass' with number 3 has low survival rate compared to other class

In[6]:

bar_survival('Sex')

It seems that male are a lot more prone to death, according to the following plt

In[7]:

bar_survival('Parch')

There aren't any dramatic difference in this feature

In[8]:

bar_survival('SibSp')

this plt also gives no insight

In[9]:

bar_survival('Age')

it seems very young childereb and very old people are prone to survival. I should use this feature to make my prediction

In[]:

​