타이타닉 캐글

[10기_임수정]

########TRAIN

import  pandas  as  pd

import  seaborn   as  sns

import  matplotlib.pyplot   as  plt 

df = pd.read_csv("d:\\data\\train.csv")

pd.set_option('display.max_columns',15)

rdf = df.drop(['Cabin'],axis=1)

# 운임 이상치 제거 

local_std = rdf.Fare.std()*5

rdf['Fare'] = rdf['Fare'][rdf['Fare']<local_std]

#rdf = rdf.dropna(subset=['Age'], how='any',axis=0)

# 나이 na값 31 로 

most_freq =rdf['Age'].value_counts(dropna=True).idxmax()

rdf['Age'].fillna(31,inplace=True)

# embarked 최빈값으로

most_freq =rdf['Embarked'].value_counts(dropna=True).idxmax()

rdf['Embarked'].fillna(most_freq,inplace=True)

# 운임 최빈값으로

most_freq =rdf['Fare'].value_counts(dropna=True).idxmax()

rdf['Fare'].fillna(most_freq,inplace=True)

ndf= rdf[['Survived','Pclass','Sex','Age','SibSp','Parch','Embarked','Fare']]

# child 와 woman 파생변수 생성

mask = ( ndf.Age < 10) | (ndf.Sex == 'female')

mask2 = (ndf)

ndf['child_women'] = mask.astype(int)

### 혼자 여행온 사람

ndf['relatives'] = ndf['SibSp'] + ndf['Parch']

ndf.loc[ndf['relatives'] > 0, 'travelled_alone'] = 0

ndf.loc[ndf['relatives'] == 0, 'travelled_alone'] = 1

###FATHER 아빠는 가족들 챙기느라 죽었을것 같음

mask1 = ( ndf.relatives > 4) | (ndf.Sex == 'male')

ndf['Father'] = mask1.astype(int)

## 더미 생성

gender=pd.get_dummies(ndf['Sex'])

ndf=pd.concat([ndf,gender],axis=1)

onehot_embarked=pd.get_dummies(ndf['Embarked'],prefix='town')

ndf=pd.concat([ndf,onehot_embarked],axis=1)

ndf.drop(['Sex','Embarked'],axis=1,inplace=True)

X=ndf[['Pclass','Father','travelled_alone', 'Age','Fare','female', 'male','Fare','town_C', 'town_Q', 'town_S','child_women']]

y=ndf['Survived']

######TEST#########

import  pandas  as  pd

import  seaborn   as  sns

import  matplotlib.pyplot   as  plt 

df1 = pd.read_csv("d:\\data\\test.csv")

pd.set_option('display.max_columns',15)

rdf = df1.drop(['Cabin'],axis=1)

most_freq =rdf['Embarked'].value_counts(dropna=True).idxmax()

rdf['Embarked'].fillna(most_freq,inplace=True)

most_freq =rdf['Fare'].value_counts(dropna=True).idxmax()

rdf['Fare'].fillna(most_freq,inplace=True)

most_freq =rdf['Age'].value_counts(dropna=True).idxmax()

rdf['Age'].fillna(31,inplace=True)

ndf= rdf[['Pclass','Sex','Age','SibSp','Parch','Embarked','Fare']]

mask = ( ndf.Age < 10) | (ndf.Sex == 'female')

ndf['child_women'] = mask.astype(int)

#relatives

ndf['relatives'] = ndf['SibSp'] + ndf['Parch']

ndf.loc[ndf['relatives'] > 0, 'travelled_alone'] = 1

ndf.loc[ndf['relatives'] == 0, 'travelled_alone'] = 0

##FATHER

mask1 = ( ndf.relatives > 4) | (ndf.Sex == 'male')

ndf['Father'] = mask1.astype(int)

gender=pd.get_dummies(ndf['Sex'])

ndf=pd.concat([ndf,gender],axis=1)

onehot_embarked=pd.get_dummies(ndf['Embarked'],prefix='town')

ndf=pd.concat([ndf,onehot_embarked],axis=1)

ndf.drop(['Sex','Embarked'],axis=1,inplace=True)

test=ndf[['Pclass','SibSp', 'Parch', 'Age','travelled_alone','Fare','female', 'male','Fare','town_C', 'town_Q', 'town_S','child_women']]

from sklearn import preprocessing 

test=preprocessing.StandardScaler().fit(test).transform(test) 

from sklearn import preprocessing 

X=preprocessing.StandardScaler().fit(X).transform(X) 

from sklearn.model_selection import train_test_split

X_train,X_test,y_train,y_test = train_test_split(X,y,test_size=0.3,random_state=19)

print('train data 의 갯수:',X_train.shape) # (478,9)

print('test data의 갯수:',X_test.shape) # (205,9)

from  sklearn.ensemble   import  RandomForestClassifier 

tree_model = RandomForestClassifier( n_estimators=800,

                                     oob_score=True,

                                    random_state= 9 )  

tree_model.fit( X, y )

# 7단계 테스트 데이터로 예측을 한다.

y_hat = tree_model.predict( test )

print ( tree_model.oob_score_)

0.8204264870931538

제가 뽑은 파생변수는 Father 와 trabelled alone 입니다 ^^

   Pclass  SibSp  Father  Parch   Age  travelled_alone     Fare  female  male  \

0       3      0       1      0  34.5              0.0   7.8292       0     1   

1       3      1       0      0  47.0              1.0   7.0000       1     0   

2       2      0       1      0  62.0              0.0   9.6875       0     1   

3       3      0       1      0  27.0              0.0   8.6625       0     1   

4       3      1       0      1  22.0              1.0  12.2875       1     0   

   town_C  town_Q  town_S  child_women  

0       0       1       0            0  

1       0       0       1            1  

2       0       1       0            0  

3       0       0       1            0  

4       0       0       1            1

제 캐글 점수는 0.75598 에요 ^^

 뭐가 문제일까요?

[유연수]

수정아 테스트 데이터의 나이의 결측치를 21로 갱신하는 코드가 없구나

[10기_임수정]

수정해도 0.75598 입니다 ㅠㅠ