sns.countplot(x='Age', data = t,hue = 'Sex')
.py
sns.countplot(x='Age', data = t,hue = 'Survived')
.py
5.まとめ
以上からどのような人が生き残りやすいのかというと15歳以下の子供あるいは女性であればかなり助かります。また、それで1等に乗っていて3人か4人家族だとほぼ生き残れるでしょう。逆に、私のような男でお金のない学生で一人旅が好きな奴は死にます。私は乗らなくてよかったです。タイタニックのおかげで救命ボートがたくさん積まれるようになり、無線も常備されたりなどして今の航海の安全に繋がっているそうです。
6.感想
とりあえずKaggleをやってみて実際にどのように機械学習をやって行けば良いのかわかった。そのためにカーネルが非常に役に立つこともわかった。今後他のKaggleの問題にも挑戦していく。この記事を書いたことによって今まではただいい結果を出すことに集中しているだけでデータの意味とかを考えなかったが今後はそれらを考慮に入れつつデータ分析を行って行きたい。また、実際はそもそもデータの加工を自分でしなければならずこのように綺麗なデータが与えられることは少ないのでSQLを勉強したい。そして、今度はHousePrices問題をやりたい。
7.用いたコード
import pandas as pd import numpy as np from sklearn.ensemble import RandomForestClassifier train= pd.read_csv("train.csv").replace("male",0).replace("female",1).replace("S",0).replace("C",1).replace("Q",2) test= pd.read_csv("test.csv").replace("male",0).replace("female",1).replace("S",0).replace("C",1).replace("Q",2) train["Age"].fillna(train.Age.mean(), inplace=True) train["Embarked"].fillna(train.Embarked.mean(), inplace=True) combine1 = [train] for train in combine1: train['Salutation'] = train.Name.str.extract(' ([A-Za-z]+).', expand=False) for train in combine1: train['Salutation'] = train['Salutation'].replace(['Lady', 'Countess','Capt', 'Col','Don', 'Dr', 'Major', 'Rev', 'Sir', 'Jonkheer', 'Dona'], 'Rare') train['Salutation'] = train['Salutation'].replace('Mlle', 'Miss') train['Salutation'] = train['Salutation'].replace('Ms', 'Miss') train['Salutation'] = train['Salutation'].replace('Mme', 'Mrs') del train['Name'] Salutation_mapping = {"Mr": 1, "Miss": 2, "Mrs": 3, "Master": 4, "Rare": 5} for train in combine1: train['Salutation'] = train['Salutation'].map(Salutation_mapping) train['Salutation'] = train['Salutation'].fillna(0) for train in combine1: train['Ticket_Lett'] = train['Ticket'].apply(lambda x: str(x)[0]) train['Ticket_Lett'] = train['Ticket_Lett'].apply(lambda x: str(x)) train['Ticket_Lett'] = np.where((train['Ticket_Lett']).isin(['1', '2', '3', 'S', 'P', 'C', 'A']), train['Ticket_Lett'], np.where((train['Ticket_Lett']).isin(['W', '4', '7', '6', 'L', '5', '8']), '0','0')) train['Ticket_Len'] = train['Ticket'].apply(lambda x: len(x)) del train['Ticket'] train['Ticket_Lett']=train['Ticket_Lett'].replace("1",1).replace("2",2).replace("3",3).replace("0",0).replace("S",3).replace("P",0).replace("C",3).replace("A",3) for train in combine1: train['Cabin_Lett'] = train['Cabin'].apply(lambda x: str(x)[0]) train['Cabin_Lett'] = train['Cabin_Lett'].apply(lambda x: str(x)) train['Cabin_Lett'] = np.where((train['Cabin_Lett']).isin([ 'F', 'E', 'D', 'C', 'B', 'A']),train['Cabin_Lett'], np.where((train['Cabin_Lett']).isin(['W', '4', '7', '6', 'L', '5', '8']), '0','0')) del train['Cabin'] train['Cabin_Lett']=train['Cabin_Lett'].replace("A",1).replace("B",2).replace("C",1).replace("0",0).replace("D",2).replace("E",2).replace("F",1) train["FamilySize"] = train["SibSp"] + train["Parch"] + 1 for train in combine1: train['IsAlone'] = 0 train.loc[train['FamilySize'] == 1, 'IsAlone'] = 1 train_data = train.values xs = train_data[:, 2:] # Pclass以降の変数 y = train_data[:, 1] # 正解データ test["Age"].fillna(train.Age.mean(), inplace=True) test["Fare"].fillna(train.Fare.mean(), inplace=True) combine = [test] for test in combine: test['Salutation'] = test.Name.str.extract(' ([A-Za-z]+)\.', expand=False) for test in combine: test['Salutation'] = test['Salutation'].replace(['Lady', 'Countess','Capt', 'Col',\ 'Don', 'Dr', 'Major', 'Rev', 'Sir', 'Jonkheer', 'Dona'], 'Rare') test['Salutation'] = test['Salutation'].replace('Mlle', 'Miss') test['Salutation'] = test['Salutation'].replace('Ms', 'Miss') test['Salutation'] = test['Salutation'].replace('Mme', 'Mrs') del test['Name'] Salutation_mapping = {"Mr": 1, "Miss": 2, "Mrs": 3, "Master": 4, "Rare": 5} for test in combine: test['Salutation'] = test['Salutation'].map(Salutation_mapping) test['Salutation'] = test['Salutation'].fillna(0) for test in combine: test['Ticket_Lett'] = test['Ticket'].apply(lambda x: str(x)[0]) test['Ticket_Lett'] = test['Ticket_Lett'].apply(lambda x: str(x)) test['Ticket_Lett'] = np.where((test['Ticket_Lett']).isin(['1', '2', '3', 'S', 'P', 'C', 'A']), test['Ticket_Lett'], np.where((test['Ticket_Lett']).isin(['W', '4', '7', '6', 'L', '5', '8']), '0', '0')) test['Ticket_Len'] = test['Ticket'].apply(lambda x: len(x)) del test['Ticket'] test['Ticket_Lett']=test['Ticket_Lett'].replace("1",1).replace("2",2).replace("3",3).replace("0",0).replace("S",3).replace("P",0).replace("C",3).replace("A",3) for test in combine: test['Cabin_Lett'] = test['Cabin'].apply(lambda x: str(x)[0]) test['Cabin_Lett'] = test['Cabin_Lett'].apply(lambda x: str(x)) test['Cabin_Lett'] = np.where((test['Cabin_Lett']).isin(['T', 'H', 'G', 'F', 'E', 'D', 'C', 'B', 'A']),test['Cabin_Lett'], np.where((test['Cabin_Lett']).isin(['W', '4', '7', '6', 'L', '5', '8']), '0','0')) del test['Cabin'] test['Cabin_Lett']=test['Cabin_Lett'].replace("A",1).replace("B",2).replace("C",1).replace("0",0).replace("D",2).replace("E",2).replace("F",1).replace("G",1) test["FamilySize"] = train["SibSp"] + train["Parch"] + 1 for test in combine: test['IsAlone'] = 0 test.loc[test['FamilySize'] == 1, 'IsAlone'] = 1 test_data = test.values xs_test = test_data[:, 1:] from sklearn.ensemble import RandomForestClassifier random_forest=RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini', max_depth=25, max_features='auto', max_leaf_nodes=None, min_samples_leaf=1, min_samples_split=15, min_weight_fraction_leaf=0.0, n_estimators=51, n_jobs=4, oob_score=False, random_state=0, verbose=0, warm_start=False) random_forest.fit(xs, y) Y_pred = random_forest.predict(xs_test) """random_forest.score(xs,y) acc_random_forest = round(random_forest.score(xs,y) * 100, 2) acc_random_forest """ import csv with open("predict_result_data5.csv", "w") as f: writer = csv.writer(f, lineterminator='\n') writer.writerow(["PassengerId", "Survived"]) for pid, survived in zip(test_data[:,0].astype(int), Y_pred.astype(int)): writer.writerow([pid, survived])
.py
参考文献
https://ja.wikipedia.org/wiki/%E3%82%BF%E3%82%A4%E3%82%BF%E3%83%8B%E3%83%83%E3%82%AF%E5%8F%B7%E6%B2%88%E6%B2%A1%E4%BA%8B%E6%95%85
https://www.kaggle.com/yassineghouzam/titanic-top-4-with-ensemble-modeling
https://www.kaggle.com/jamsterjai/pythanic-please-critique-0-79904
https://www.kaggle.com/arthurtok/introduction-to-ensembling-stacking-in-python
https://www.kaggle.com/headsortails/pytanic
http://www.statisticalconsultants.co.nz/blog/titanic-survival-data.html
http://www.mirandora.com/?p=1804
http://blog.tatsushim.com/?p=63
Takumi Ihara
