Model selection

通常机器学习最难的是选择合适的estimator，不同的estimator适用于不同的数据集和问题

def gen_estimators():

    '''

    List of the different estimators.

    '''

    estimators = [

        ('Lasso regression',linear_model.Lasso(alpha=0.1),True),

        ('Ridge regression',linear_model.Ridge(alpha=0.1),True),

        ('Hinge regression',linear_model.Hinge(),True),

        ('Lassolars regression',linear_model.LassoLars(alpha=0.1),True),

        ('OrthogonalMatchingPursuitCV regression',linear_model.OrthogonalMatchingPursuitCV(),True),

        ('BayesianRidge regression',linear_model.BayesianRidge(),True),

        ('PassiveAggressiveRegressor regression',linear_model.PassiveAggressiveRegressor(),True),

        ('HuberRegressor regression',linear_model.Huber(),True),

        ('LogisticRegression regression',linear_model.LogisticRegression(),True),

    ]

    return estimators

遍历，观察得分

def cross_validate():

    for name,clf,flag in gen_estimators():

        x_train,x_test,y_train,y_test = train_test_split(X,y,test_size=0.4,random_state=0)

        clf.fit(x_train,y_train)

        print(name,'\n',clf.coef_)

        scores = cross_val_score(clf,X,y,cv=5,scoring='roc_aus')

        print(scores)

scikit-learn交叉验证

hold测试：训练集和测试集分割

from sklearn.datasets import load_iris

from sklearn.cross_validation import train_test_split

iris = load_iris()

X = iris.data

y = iris.target

x_train,x_test,y_train,y_test = train_test_split(X,y,test_size=0.4,random_state=0) #随机数种子，没有设置，得到的划分每次都不一样

print(y_train)

print(y_test)