1 # 接下来我们将一些变化幅度较大的特征化到[-1,1]之内，这样可以加速logistic regression的收敛
2 import sklearn.preprocessing as preprocessing
3 scaler = preprocessing.StandardScaler()
4 age_scale_param = scaler.fit(df['Age'])
5 df['Age_scaled'] = scaler.fit_transform(df['Age'],age_scale_param)
6 fare_scale_param = scaler.fit(df['Fare'])
7 df['Fare_scaled'] = scaler.fit_transform(df['Fare'],fare_scale_param)
8 print(df)

1 from sklearn import linear_model
2 train_df = df.filter(regex='Survived|Age_.*|SibSp|Parch|Fare_.*|Cabin_.*|Embarked_.*|Sex_.*|Pclass_.*')
3 train_np = train_df.as_matrix()
4 y = train_np[:,0]
5 X = train_np[:,1:]
6 clf = linear_model.LogisticRegression(C=1.0,penalty='l1',tol=1e-6)
7 clf.fit(X,y)
8 print(clf)

 1 # #首先用同样的RandomForestRegressor模型填上丢失的年龄
 2 data_test = pd.read_csv("test.csv")
 3 data_test.loc[ (data_test.Fare.isnull()), 'Fare' ] = 0
 4
 5 tmp_df = data_test[['Age','Fare', 'Parch', 'SibSp', 'Pclass']]
 6 null_age = tmp_df[data_test.Age.isnull()].as_matrix()
 7 # 根据特征属性X预测年龄并补上
 8 X = null_age[:, 1:]
 9 predictedAges = rfr.predict(X)
10 data_test.loc[ (data_test.Age.isnull()), 'Age' ] = predictedAges
11
12 data_test = set_Cabin_type(data_test)
13 dummies_Cabin = pd.get_dummies(data_test['Cabin'], prefix= 'Cabin')
14 dummies_Embarked = pd.get_dummies(data_test['Embarked'], prefix= 'Embarked')
15 dummies_Sex = pd.get_dummies(data_test['Sex'], prefix= 'Sex')
16 dummies_Pclass = pd.get_dummies(data_test['Pclass'], prefix= 'Pclass')
17
18
19 df_test = pd.concat([data_test, dummies_Cabin, dummies_Embarked, dummies_Sex, dummies_Pclass], axis=1)
20 df_test.drop(['Pclass', 'Name', 'Sex', 'Ticket', 'Cabin', 'Embarked'], axis=1, inplace=True)
21 df_test['Age_scaled'] = scaler.fit_transform(df_test['Age'], age_scale_param)
22 df_test['Fare_scaled'] = scaler.fit_transform(df_test['Fare'], fare_scale_param)
23
24 test = df_test.filter(regex='Age_.*|SibSp|Parch|Fare_.*|Cabin_.*|Embarked_.*|Sex_.*|Pclass_.*')
25 predictions = clf.predict(test)
26 result = pd.DataFrame({'PassengerId':data_test['PassengerId'].as_matrix(), 'Survived':predictions.astype(np.int32)})
27 result.to_csv("logistic_regression_predictions.csv", index=False)

 1 import numpy as np
 2 import matplotlib.pyplot as plt
 3 from sklearn.learning_curve import learning_curve
 4
 5 #用sklearn的learning_curve得到training_score和cv_score，使用matplotlib画出learning curve
 6
 7 def plot_learning_curve(estimator,title,X,y,ylim=None,cv=None,n_jobs=1,train_sizes=np.linspace(.05,1.,20),verbose=0,plot=True):
 8     """
 9        画出data在某模型上的learning curve.
10        参数解释
11        ----------
12        estimator : 你用的分类器。
13        title : 表格的标题。
14        X : 输入的feature，numpy类型
15        y : 输入的target vector
16        ylim : tuple格式的(ymin, ymax), 设定图像中纵坐标的最低点和最高点
17        cv : 做cross-validation的时候，数据分成的份数，其中一份作为cv集，其余n-1份作为training(默认为3份)
18        n_jobs : 并行的的任务数(默认1)
19        """
20     train_sizes, train_scores, test_scores = learning_curve(
21         estimator, X, y, cv=cv, n_jobs=n_jobs, train_sizes=train_sizes, verbose=verbose)
22
23     train_scores_mean = np.mean(train_scores, axis=1)
24     train_scores_std = np.std(train_scores, axis=1)
25     test_scores_mean = np.mean(test_scores, axis=1)
26     test_scores_std = np.std(test_scores, axis=1)
27
28     if plot:
29         plt.figure()
30         plt.title(title)
31         if ylim is not None:
32             plt.ylim(*ylim)
33         plt.xlabel('train_sample')
34         plt.ylabel('score')
35         plt.gca().invert_yaxis()
36         plt.grid()
37
38         plt.fill_between(train_sizes, train_scores_mean - train_scores_std, train_scores_mean + train_scores_std,
39                          alpha=0.1, color="b")
40         plt.fill_between(train_sizes, test_scores_mean - test_scores_std, test_scores_mean + test_scores_std,
41                          alpha=0.1, color="r")
42         plt.plot(train_sizes, train_scores_mean, 'o-', color="b", label="train_score")
43         plt.plot(train_sizes, test_scores_mean, 'o-', color="r", label="cross_validation_score")
44
45         plt.legend(loc='best')
46         plt.draw()
47         plt.gca().invert_yaxis()
48         plt.show()
49
50     midpoint = ((train_scores_mean[-1] + train_scores_std[-1]) + (test_scores_mean[-1] - test_scores_std[-1])) / 2
51     diff = (train_scores_mean[-1] + train_scores_std[-1]) - (test_scores_mean[-1] - test_scores_std[-1])
52     return midpoint, diff
53
54 plot_learning_curve(clf, u"学习曲线", X, y)