用于 svm 的机器学习网格搜索

Question

我正在做一个项目，我需要计算网格搜索返回的最佳估计器。

parameters = {'gamma':[0.1, 0.5, 1, 10, 100], 'C':[1, 5, 10, 100, 1000]}
# TODO: Initialize the classifier
svr = svm.SVC()
# TODO: Make an f1 scoring function using 'make_scorer' 
f1_scorer = make_scorer(score_func)
# TODO: Perform grid search on the classifier using the f1_scorer as the scoring method
grid_obj = grid_search.GridSearchCV(svr, parameters, scoring=f1_scorer)
# TODO: Fit the grid search object to the training data and find the optimal parameters
grid_obj = grid_obj.fit(X_train, y_train)
pred = grid_obj.predict(X_test)
def score_func():
    f1_score(y_test, pred, pos_label='yes')
# Get the estimator
clf = grid_obj.best_estimator_

我不确定如何使f1_scorer功能，因为我在创建网格搜索对象后进行了预测。创建 obj 后我无法声明f1_scorer，因为网格搜索将其用作评分方法。请帮助我如何为网格搜索创建此评分函数。

Answer 1

clf = svm.SVC()
# TODO: Make an f1 scoring function using 'make_scorer' 
f1_scorer = make_scorer(f1_score,pos_label='yes')
# TODO: Perform grid search on the classifier using the f1_scorer as the scoring method
grid_obj = GridSearchCV(clf,parameters,scoring=f1_scorer)
# TODO: Fit the grid search object to the training data and find the optimal parameters
grid_obj = grid_obj.fit(X_train, y_train)
# Get the estimator
clf = grid_obj.best_estimator_

Answer 2

您传递给make_scorer的记分器函数应该采用y_true并y_pred为参数。有了这些信息，你就有了计算分数所需的一切。然后 GridSearchCV 将针对每个可能的参数集在内部为您拟合并调用 score 函数，您无需事先计算y_pred。

它应该看起来像这样：

def score_func(y_true, y_pred):
    """Calculate f1 score given the predicted and expected labels"""
    return f1_score(y_true, y_pred, pos_label='yes')
f1_scorer = make_scorer(score_func)
GridSearchCV(svr, parameters, scoring=f1_scorer)