def do_xgboost_blance_sample(X, Y):
"""
Args:
X_train:DataFrame, shape(x_index*(1-test_size), feature_number(feature_map)), the features for training
X_test: DataFrame, shape(x_index*test_size, feature_number(feature_map)), the features for testing
Y_train: list, the labels for training
Y_test: list, the labels for testing
Returns:
Y_pred: list, the Predicted value of the model
"""
logging.info('训练结果输出')
X_train, X_vld, Y_train, Y_vld = train_test_split(X,
Y,
test_size=0.2,
random_state=1)
xgb_model = xgb.XGBClassifier().fit(X_train,
Y_train,
sample_weight=compute_sample_weight(
"balanced", Y_train))
Y_pred = xgb_model.predict(X_vld)
do_metrics(Y_vld, Y_pred)
return xgb_model
def dnn_test(X_test, Y_test, model_path, log_path, name):
"""
Args:
X_test:
Y_test:
save_path:
name:
Returns:
"""
save_log_file(log_path)
model = load_model(model_path + name + 'model.h5')
Y_test = to_categorical(Y_test, num_classes=5)
Y_pred = model.predict(X_test)
do_metrics(Y_test, Y_pred)
attack_types = ['normal', 'attacker', 'victim', 'suspicious', 'unknown']
confusion_matrixs.plot_confusion_matrix(
np.array(metrics.confusion_matrix(Y_test, Y_pred)),
classes=attack_types,
normalize=True,
title='dnn Normalized confusion matrix')
if __name__=="__main__":
log_path = '/home/liyulian/code/CIDDS/repositories'
save_log_file(log_path)
### CIDDS-001
path = '/home/liyulian/code/CIDDS/sources/utils/data_features_001.csv'
### 无样本平衡
logging.info('使用CIDDS-001数据, 用xgboost完成实验分类,'
'无样本平衡, 设置了划分样本的random_state=1, 并且保存模型')
X_train, X_test, Y_train, Y_test = get_features_FPPB(path)
Y_pred, model = machine_learnings.do_xgboost(X_train, X_test, Y_train, Y_test)
do_metrics(Y_test, Y_pred)
import pickle # pickle模块
# 保存Model(注:save文件夹要预先建立，否则会报错)
with open('repositories/model_001.pickle', 'wb') as f:
pickle.dump(model, f)
# ### 无样本平衡
# logging.info('使用CIDDS-001数据, 用xgboost完成实验分类,'
# '无样本平衡')
#
# X_train, X_test, Y_train, Y_test = get_features(path)
# Y_pred = machine_learnings.do_xgboost(X_train, X_test, Y_train, Y_test)
# do_metrics(Y_test, Y_pred)
"""