如何遍歷不同的 sci-kit 學習分類器

Question

我正在使用 scikit-learn 運行一堆模型來解決分類問題。

如何迭代不同的 scikit-learn 模型？

from sklearn.ensemble import AdaBoostClassifier
from sklearn.naive_bayes import BernoulliNB
from sklearn.dummy import DummyClassifier

classifiers_name = ['AdaBoostClassifier',
                    'BernoulliNB',
                    'DummyClassifier']

def fitting_classifier(clf, X_train, y_train):
    return clf.fit(X_train, y_train)

for clf_n in classifiers_name:
    locals()['results_' + clf_n] = fitting_classifier(locals()[clf_n + str(())], X_train, y_train)

我似乎在這部分代碼中遇到了錯誤： fitting_classifier(locals()[clf_n + str(())], X_train, y_train) 。 顯示的錯誤是：

<ipython-input-31-cccf30ff4392> in summary_scores(file_path, image_format, scores)
    140         for clf_sn in classifiers_name:
--> 141             locals()['results_' + clf_n] = fitting_classifier(locals()[clf_n + str(())], X_train, y_train)
    142 
    143         # results_AdaBoostClassifier = fitting_classifier(AdaBoostClassifier(), X_train, y_train)

KeyError: 'AdaBoostClassifier()'

對此的任何幫助將不勝感激。 謝謝你。

Answer 1

由於您沒有提到這樣做的目的。 為什么您要迭代不同的 scikit-learn 模型？

如果您想找出上面的哪個 model 更適合並且表現更好，您可以使用類似這樣的東西

# -------- Cross validate model with Kfold stratified cross val ---------------

    kfold = StratifiedKFold(n_splits=10)

# Modeling step Test differents algorithms
    classifiers = ['AdaBoostClassifier',
                    'BernoulliNB',
                    'DummyClassifier']
    results = []
    for model in classifiers :
        results.append(cross_val_score(model, X_train, y = y_train, scoring = "accuracy", cv = kfold, n_jobs=4))

    cv_means = []
    cv_std = []
    for cv_result in results:
        cv_means.append(cv_result.mean())
        cv_std.append(cv_result.std())

    cv_res = pd.DataFrame({"CrossValMeans":cv_means,"CrossValerrors": cv_std,"Algorithm":["AdaBoostClassifier","BernoulliNB","DummyClassifier"]})`

如果你正在嘗試合奏這些

分別訓練它們，並使用 HyperParams 找到 model 的最佳估計器，然后將VotingClassifier用作：

    DTC = DecisionTreeClassifier()
    ADB = AdaBoostClassifier(DTC)

    ada_param_grid = { # Params here }

    gsABC = GridSearchCV(ADB,param_grid = ada_param_grid , cv=kfold, scoring="accuracy", n_jobs= 4, verbose = 1)

    AdaBoost_best =gsABC.best_estimator_

 # Likewise you can do for others and then perform Voting

    votingC = VotingClassifier(estimators=[('ada', AdaBoost_best), ('nb', BernoulliNB_best),
    ('dc', DummyClassifier_best)], voting='soft', n_jobs=4)

    votingC = votingC.fit(X_train, Y_train)