def callRFC():
### Feature selection
num_to_keep = 9
my_features_list = select_k_best_features(my_dataset, features_list, num_to_keep)
### Make classier pipeline
pipeline_rfc = Pipeline(steps=[
('classifier', RandomForestClassifier(random_state = 42))
])
parameters_rfc = {
'classifier__max_features': ('sqrt', 1),
'classifier__max_depth': np.arange(3, 8),
'classifier__n_estimators' : (10, 20)
}
### Grid search for the optimal parameters
precision_n_recall(pipeline_rfc, parameters_rfc, my_dataset, my_features_list)
clf = RandomForestClassifier(max_depth = 5,
max_features = 'sqrt',
n_estimators = 10,
random_state = 42)
tester_prep(clf, my_dataset, my_features_list)
def callSVC():
### Feature selection
num_to_keep = 8
my_features_list = select_k_best_features(my_dataset, features_list, num_to_keep)
### Make classier pipeline
pipeline_svc = Pipeline(steps=[
('scaler', StandardScaler()),
('classifier', SVC(kernel = 'rbf', random_state = 42, class_weight = 'auto'))
])
parameters_svc = {
'classifier__gamma': 10.0 ** np.arange(-4, 0),
'classifier__C': 10.0 ** np.arange(1, 5)
}
### Grid search for the optimal parameters
precision_n_recall(pipeline_svc, parameters_svc, my_dataset, my_features_list)
clf = Pipeline(steps=[
('scaler', StandardScaler()),
('classifier', SVC(kernel = 'rbf', C = 1000, gamma = 0.0001,
random_state = 42, class_weight = 'auto'))
])
tester_prep(clf, my_dataset, my_features_list)
def callNBC():
### Feature selection
num_to_keep = 8
my_features_list = select_k_best_features(my_dataset, features_list, num_to_keep)
clf = GaussianNB()
tester_prep(clf, my_dataset, my_features_list)
def callLR():
### Feature selection
num_to_keep = 16
my_features_list = select_k_best_features(my_dataset, features_list, num_to_keep)
### Make classier pipeline
pipeline_lrg = Pipeline(steps=[
('scaler', StandardScaler()),
('classifier', LogisticRegression(tol = 0.001, random_state = 42))
])
parameters_lrg = {
'classifier__penalty': ('l1', 'l2'),
'classifier__C': 10.0 ** np.arange(-10, -3)
}
### Grid search for the optimal parameters
precision_n_recall(pipeline_lrg, parameters_lrg, my_dataset, my_features_list)
clf = Pipeline(steps=[
('scaler', StandardScaler()),
('classifier', LogisticRegression(tol = 0.001, C = 10**-8, penalty = 'l2',
random_state = 42))
])
tester_prep(clf, my_dataset, my_features_list)
