titanic

gistfile1.txt

import sys
import csv
import pandas as pd
from pprint import pprint
from sklearn import linear_model
from sklearn import svm
from sklearn.preprocessing import StandardScaler

def read_csv(filename):
    df = pd.read_csv(filename)
    df = df.drop(['Name', 'Cabin', 'Embarked', 'Ticket'], axis=1)

    # 男:1 女: -1
    def sex_to_int(x):
        if x == 'male':
            return 1
        elif x == 'female':
            return -1
        else:
            return 0
    df['Sex'] = df['Sex'].map(sex_to_int)
    return df


def regressions():
    kw = dict()
    return [
        linear_model.LinearRegression(normalize=True),
        linear_model.LogisticRegression(),
        svm.SVC(kernel="poly", degree=2, **kw),
        svm.SVC(kernel="poly", degree=3, **kw),
        svm.SVC(kernel="poly", degree=4, **kw),
        svm.SVC(kernel="rbf", **kw),
        svm.SVC(kernel="sigmoid", **kw),
    ]
    

def main():
    scaler = StandardScaler()
    train_df = read_csv('train.csv').drop('PassengerId', axis=1)
    
    age_mean = train_df['Age'].mean()
    fare_mean = train_df['Fare']
    train_df['Age'] = train_df['Age'].fillna(age_mean)
    train_df['Fare'] = train_df['Fare'].fillna(fare_mean)
    train_df['Survived'] = train_df['Survived'].map(lambda x: 2 * x - 1)

    train_survived = train_df['Survived']
    train_df = train_df.drop('Survived', axis=1)

    train_df1 = train_df[0:800]
    train_survived1 = train_survived[0:800]
    train_df2 = train_df[800:]
    train_survived2 = train_survived[800:]

    scaler.fit(train_df1)


    fitted_regs = []
    for reg in regressions():
        print(reg)
        reg.fit(scaler.transform(train_df1), train_survived1)  
        estimated = reg.predict(scaler.transform(train_df2))

        correct = 0
        for _, item in pd.DataFrame({ 'actual': train_survived2, 'estimated': estimated }).iterrows():
            if (item['actual'] == 1) == (item['estimated'] >= 0.5):
                correct += 1
        print(correct / len(train_survived2.index))
        print()

        fitted_regs.append(reg)

    if True:
        reg = fitted_regs[5]
        print(reg)
        test_df = read_csv('test.csv')
        # Fare が nan の場合があるので、訓練データの平均値で穴埋め
        test_df['Age']  = test_df['Age'].fillna(age_mean)
        test_df['Fare'] = test_df['Fare'].fillna(fare_mean)

        ids = test_df['PassengerId']
        test_df = test_df.drop('PassengerId', axis=1)
        
        survived = reg.predict(scaler.transform(test_df))
        # print(survived[0])
        survived = [(1 if x >= 0 else 0 ) for x in survived]
        answer = pd.DataFrame({ 'PassengerId': ids, 'Survived': survived })
        answer.to_csv('answer.csv', index=False)


if __name__ == '__main__':
    main()