타이타닉 생존자 예측 대회¶

학습 목표¶

  • Cabin 피처를 사용한다.
  • 원핫 인코딩을 이해한다.
  • GridSearchCV에 대해 이해한다.

목차

01. 데이터 불러오기
02. 데이터 전처리
03. 모델 구축 및 학습

데이터¶

Data Fields¶

구분 설명 값
Survival 생존 여부 Survival. 0 = No, 1 = Yes
Pclass 티켓의 클래스 Ticket class. 1 = 1st, 2 = 2nd, 3 = 3rd
Sex 성별(Sex) 남(male)/여(female)
Age 나이(Age in years.)
SibSp 함께 탑승한 형제와 배우자의 수 /siblings, spouses aboard the Titanic.
Parch 함께 탑승한 부모, 아이의 수 # of parents / children aboard the Titanic.
Ticket 티켓 번호(Ticket number) (ex) CA 31352, A/5. 2151
Fare 탑승료(Passenger fare)
Cabin 객실 번호(Cabin number)
Embarked 탑승 항구(Port of Embarkation) C = Cherbourg, Q = Queenstown, S = Southampton
  • siblings : 형제, 자매, 형제, 의붓 형제
  • spouses : 남편, 아내 (정부와 약혼자는 무시)
  • Parch : Parent(mother, father), child(daughter, son, stepdaughter, stepson)

01. 데이터 불러오기

목차로 이동하기

참고 노트북¶

  • titanic 전체 노트북
    • https://www.kaggle.com/code/pliptor/how-am-i-doing-with-my-score/report
    • https://www.kaggle.com/code/ccastleberry/titanic-cabin-features/notebook
In [1]:
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
In [2]:
train = pd.read_csv("data/titanic/train.csv", index_col='PassengerId')
test = pd.read_csv("data/titanic/test.csv", index_col='PassengerId')
sub = pd.read_csv("data/titanic/gender_submission.csv")

데이터 합치기¶

In [3]:
train_results = train["Survived"].copy()
train.drop("Survived", axis=1, inplace=True, errors="ignore")
titanic = pd.concat([train, test])
traindex = train.index
testdex = test.index

Cabin 확인¶

In [4]:
titanic.isnull().sum()
Out[4]:
Pclass         0
Name           0
Sex            0
Age          263
SibSp          0
Parch          0
Ticket         0
Fare           1
Cabin       1014
Embarked       2
dtype: int64
In [5]:
titanic['Cabin'].value_counts()
Out[5]:
C23 C25 C27        6
G6                 5
B57 B59 B63 B66    5
C22 C26            4
F33                4
                  ..
A14                1
E63                1
E12                1
E38                1
C105               1
Name: Cabin, Length: 186, dtype: int64

02. Cabin 데이터 전처리

목차로 이동하기

Cabin 데이터 전처리¶

In [16]:
cabin_only = titanic[["Cabin"]].copy()
cabin_only["Cabin_Data"] = cabin_only["Cabin"].isnull().apply(lambda x: not x)
In [17]:
cabin_only["Deck"] = cabin_only["Cabin"].str.slice(0,1)
cabin_only["Room"] = cabin_only["Cabin"].str.slice(1,5).str.extract("([0-9]+)", 
                                                     expand=False).astype("float")
cabin_only[cabin_only["Cabin_Data"]]
Out[17]:
Cabin Cabin_Data Deck Room
PassengerId
2 C85 True C 85.0
4 C123 True C 123.0
7 E46 True E 46.0
11 G6 True G 6.0
12 C103 True C 103.0
... ... ... ... ...
1296 D40 True D 40.0
1297 D38 True D 38.0
1299 C80 True C 80.0
1303 C78 True C 78.0
1306 C105 True C 105.0

295 rows × 4 columns

In [18]:
cabin_only['Deck'].value_counts()
Out[18]:
C    94
B    65
D    46
E    41
A    22
F    21
G     5
T     1
Name: Deck, dtype: int64
In [19]:
cabin_only['Deck'].isnull().sum()
Out[19]:
1014

Cabin, Cabin_data 삭제¶

In [20]:
cabin_only.drop(["Cabin", "Cabin_Data"], axis=1, inplace=True, errors="ignore")
In [21]:
cabin_only["Deck"] = cabin_only["Deck"].fillna("N")
cabin_only["Room"] = cabin_only["Room"].fillna(cabin_only["Room"].mean())
In [22]:
cabin_only.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 1309 entries, 1 to 1309
Data columns (total 2 columns):
 #   Column  Non-Null Count  Dtype  
---  ------  --------------  -----  
 0   Deck    1309 non-null   object 
 1   Room    1309 non-null   float64
dtypes: float64(1), object(1)
memory usage: 30.7+ KB

'''
선택된 컬럼의 원핫을 수행
Args:
    df: 판다스 데이터 프레임
    label: 변경할 컬럼명
    drop_col: boolean to decide if the chosen column should be dropped
Returns:
    원핫이 추가된 데이터 프레임
'''
In [23]:
def one_hot_column(df, label, drop_col=False):
    one_hot = pd.get_dummies(df[label], prefix=label)
    if drop_col:
        df = df.drop(label, axis=1)
    df = df.join(one_hot)
    return df

'''
This function will one hot encode a list of columns.
Args:
    df: Pandas dataframe
    labels: list of the columns to encode
    drop_col: boolean to decide if the chosen column should be dropped
Returns:
    pandas dataframe with the given encoding
'''
In [24]:
def one_hot(df, labels, drop_col=False):

    for label in labels:
        df = one_hot_column(df, label, drop_col)
    return df
In [25]:
cabin_only = one_hot(cabin_only, ["Deck"],drop_col=True)
In [26]:
cabin_only.head()
Out[26]:
Room Deck_A Deck_B Deck_C Deck_D Deck_E Deck_F Deck_G Deck_N Deck_T
PassengerId
1 49.615917 0 0 0 0 0 0 0 1 0
2 85.000000 0 0 1 0 0 0 0 0 0
3 49.615917 0 0 0 0 0 0 0 1 0
4 123.000000 0 0 1 0 0 0 0 0 0
5 49.615917 0 0 0 0 0 0 0 1 0

Pclass와 Cabin 데이터 셋의 상관관계¶

In [27]:
for column in cabin_only.columns.values[1:]:
    titanic[column] = cabin_only[column]
In [28]:
titanic.drop(["Ticket","Cabin"], axis=1, inplace=True)
In [29]:
corr = titanic.corr()
In [30]:
corr["Pclass"].sort_values(ascending=False)
Out[30]:
Pclass    1.000000
Deck_N    0.713857
SibSp     0.060832
Deck_G    0.052133
Parch     0.018322
Deck_F    0.013122
Deck_T   -0.042750
Deck_A   -0.202143
Deck_E   -0.225649
Deck_D   -0.265341
Deck_B   -0.353414
Age      -0.408106
Deck_C   -0.430044
Fare     -0.558629
Name: Pclass, dtype: float64
In [31]:
# Train
train_df = cabin_only.loc[traindex, :]
train_df['Survived'] = train_results

# Test
test_df = cabin_only.loc[testdex, :]
In [32]:
test_df.head()
Out[32]:
Room Deck_A Deck_B Deck_C Deck_D Deck_E Deck_F Deck_G Deck_N Deck_T
PassengerId
892 49.615917 0 0 0 0 0 0 0 1 0
893 49.615917 0 0 0 0 0 0 0 1 0
894 49.615917 0 0 0 0 0 0 0 1 0
895 49.615917 0 0 0 0 0 0 0 1 0
896 49.615917 0 0 0 0 0 0 0 1 0

03. 모델 구축 및 모델 평가

목차로 이동하기

In [33]:
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import RandomizedSearchCV
from sklearn import metrics
from sklearn.model_selection import cross_val_score
import scipy.stats as st
import numpy as np
In [34]:
model = RandomForestClassifier()
In [35]:
X = train_df.drop("Survived", axis=1).copy()
y = train_df["Survived"]
In [36]:
param_grid ={'max_depth': st.randint(6, 11),
             'n_estimators':st.randint(300, 500),
             'max_features':np.arange(0.5,.81, 0.05),
             'max_leaf_nodes':st.randint(6, 10)}

grid = RandomizedSearchCV(model,
                    param_grid, cv=10,
                    scoring='accuracy',
                    verbose=1,n_iter=20)

grid.fit(X, y)

Fitting 10 folds for each of 20 candidates, totalling 200 fits
Out[36]:
RandomizedSearchCV(cv=10, estimator=RandomForestClassifier(), n_iter=20,
                   param_distributions={'max_depth': <scipy.stats._distn_infrastructure.rv_frozen object at 0x000002571D175C40>,
                                        'max_features': array([0.5 , 0.55, 0.6 , 0.65, 0.7 , 0.75, 0.8 ]),
                                        'max_leaf_nodes': <scipy.stats._distn_infrastructure.rv_frozen object at 0x000002571C83D700>,
                                        'n_estimators': <scipy.stats._distn_infrastructure.rv_frozen object at 0x000002571CA05B20>},
                   scoring='accuracy', verbose=1)
In [37]:
grid.best_estimator_
Out[37]:
RandomForestClassifier(max_depth=6, max_features=0.5, max_leaf_nodes=6,
                       n_estimators=389)
In [38]:
grid.best_score_
Out[38]:
0.690274656679151
In [39]:
pred = grid.best_estimator_.predict(test_df)
In [45]:
results_df = pd.DataFrame()
results_df["PassengerId"] = test_df.index
results_df["Survived"] = pred
In [46]:
results_df.head()
Out[46]:
PassengerId Survived
0 892 0
1 893 0
2 894 0
3 895 0
4 896 0
In [47]:
results_df.to_csv("pred_2211.csv", index=False)

0.67703¶

실습¶

    1. 다른 feature도 추가한 이후에 제출해 보기
      • 'PassengerId', 'Pclass', 'SibSp', 'Parch'
    1. Age, Fare, Sex, Embarked를 추가한 이후에 제출해 보기