kaggle Titanic Tutorial – 5

ファミリーサイズを追加してみる。

ソースはこちらを参照した。

https://www.kaggle.com/lperez/titanic-a-deeper-look-on-family-size

主要な変更部分だけ記載

まずはfamilysizeを作成する

d_train = pd.read_csv('train.csv')
d_test = pd.read_csv('test.csv')

d_train['title'] = d_train['Name'].apply(get_title).map(Title_Dictionary)
d_test['title'] = d_test['Name'].apply(get_title).map(Title_Dictionary)
d_train['FamilySize'] = d_train['SibSp'] + d_train['Parch'] + 1
d_test['FamilySize'] = d_test['SibSp'] + d_test['Parch'] + 1

ParchとSibSpはどろっぷ

d_train = d_train.drop(['PassengerId','Name','Ticket','Cabin', 'Parch','SibSp'], axis=1)

分析にFamilySizeを追加、ParchとSibSpを削除する。

x_train = d_train[["Pclass", "Sex", "Age", "Fare", "Embarked", "FamilySize", "title"]].values
x_test = d_test[["Pclass", "Sex", "Age", "Fare", "Embarked", "FamilySize", "title"]].values

from sklearn.tree import DecisionTreeClassifier
dtree = DecisionTreeClassifier(max_depth=8)
dtree.fit(x_train,y_train)

predictions = dtree.predict(x_test)

ようやく6000番台に突入。