poi_id.py

features_train, features_test, labels_train, labels_test = train_test_split(features, labels, test_size=0.3,
                                                                            random_state=42)

clf = DecisionTreeClassifier(random_state=RANDOM_STATE)


param_grid = [
    {
        "pca__n_components": range(1, len(features_list) - 1, 1),
        "selectatmostkbest__k": [2, 3, 4, 5, 6, 'all'],
        "decisiontreeclassifier__criterion": ['gini', 'entropy'],
        "decisiontreeclassifier__splitter": ['best', 'random'],
        "decisiontreeclassifier__presort": [True, False]

    }
]

pipe = make_pipeline(StandardScaler(), PCA(n_components=len(features_list) - 1),
                             SelectAtMostKBest(k=len(features_list) - 1), clf)
cv = StratifiedShuffleSplit()
search = GridSearchCV(pipe, params, cv=cv, scoring=score, n_jobs=-1)
search.fit(features_train, labels_train)
search.score(features_test, labels_test)

print("BEST SCORE = {}, CLF = {}".format(str(search.best_score_), clf))
print("BEST PARAMS = " + str(search.best_params_))
print("BEST ESTIMATOR = " + str(search.best_estimator_))