# 'etc__n_estimators': [20, 200],
# 'voting': 'hard',
# If we already have the optimal config or decide not to
# fine tune each classifier individually, then
# probably try to use 'soft' voting method. However, to use soft
# voting, the weights for each classifier need to setup
# correctly. Then find the right weight combination will be the
# next target.
'voting': ('soft', ),
'weights': [[1, 1, 1], [1, 1, 2], [1, 2, 1], [2, 1, 1], [1, 2,
2], [2, 1, 2],
[2, 2, 1], [1, 2, 3], [1, 3, 2], [2, 1, 3], [2, 3, 1],
[3, 1, 2], [3, 2, 1]]
}
scorer = cbs.scorer(show=True)
if __name__ == '__main__':
# n_jobs = 1 due to limitied memory
# refit = False for the sake of clarity
gscv = GridSearchCV(vclf,
parameters,
scoring=scorer,
cv=3,
verbose=10,
refit=False,
n_jobs=1,
return_train_score=False)
gscv.fit(X, y)
print(gscv.best_params_, gscv.best_score_)
print(gscv.cv_results_)
selected_feat_names = pickle.load(f)
print("data loaded")
y = df["attack_type"].values
X = df[selected_feat_names].values
etc = ExtraTreesClassifier(n_jobs=-1)
parameters_etc = {
# 'criterion': ("gini", "entropy"),
'n_estimators': (5, 10, 20, 30, 40, 50),
# 'max_features': ("sqrt", "log2"),
# 'min_samples_split': (2, 3, 4),
}
scorer = cbs.scorer(True)
if __name__ == '__main__':
gscv = GridSearchCV(estimator=etc,
param_grid=parameters_etc,
scoring=scorer,
verbose=10,
refit=False,
cv=3,
n_jobs=1,
return_train_score=False)
gscv.fit(X, y)
print(gscv.cv_results_)
print(gscv.best_params_)
print(gscv.best_score_)
# 5 estimator, entropy, others no significance
df = pd.read_csv(r'data/train10pc', header=None, names=__ATTR_NAMES)
# sparse feature merge
df = processing.merge_sparse_feature(df)
# one hot encoding
df = processing.one_hot(df)
# y labels mapping
df = processing.map2major5(df)
with open(r'data/selected_feat_names.pkl', 'rb') as f:
selected_feat_names = pickle.load(f)
print("training data loaded")
y = df["attack_type"].values
X = df[selected_feat_names].values
ada = AdaBoostClassifier()
parameters = {
'n_estimators': (50, 75, 100, 125, 150),
'learning_rate': (1, 1.5, 2),
}
scorer = cost_based_scoring.scorer(show=True)
gscv = GridSearchCV(ada,
parameters,
scoring=scorer,
cv=3,
verbose=2,
refit=False,
n_jobs=1,
return_train_score=False)
gscv.fit(X, y)
print("optimization params:", gscv.best_params_)
print("grid search finished")
#n_estimators=75, learning_rate=1.5