def detect(self):
self.type.detector()
type1_result = self.type.type1_pairs
type2_result = self.type.similar_pairs
utility.write_result("./type1.dup", type1_result)
utility.write_result("./type2.dup", type2_result)
else:
data_files, add_to_name, data_config = common_code()
algo_name = algo_core+algo_type+"MultiStep"+str(multistep) +"Window"+\
str(window_size)+anomalyScore_func+anomalyScore_type+add_to_name
model = predictionLstmStepAhead(input_shape, multistep)
result_files = use_whole_data(data_files,
input_shape,
train_nStepPrediction_based_models_new,
model,
nStepAhead=multistep,
nb_epoch=nb_epoch,
anomaly_score=anomalyScore_type)
print(algo_name)
write_result(algorithm_name=algo_name,
data_files=result_files,
results_path=cwd + '/results')
store_param(window_size, nb_epoch, input_shape, algo_core, algo_type,
algo_name, model, normalized_input, anomalyScore_func,
anomalyScore_type, multistep)
#for i in range(len(df)):
# a.append(al.anomalyProbability(df.value.values[i],df.anomaly_score.values[i],df.timestamp.values[i]))
# 1- Params of model
# 2- Params of training
# 3- Get model
# 4- Get type of training
# 5- Train and get result
# 6- Write output and params
from keras.layers import Conv1D, Flatten, Dropout, Dense
from utility import read_data, train_autoencoder_based_models, use_whole_data, write_result, common_code_normalized
from models import autoencoderLstm
import os
import pickle
from datetime import datetime
now = datetime.now()
cwd = os.getcwd()
path = cwd + "/data"
data_files = read_data(path)
window_size = 10
nb_epoch = 20
nb_features = 1
input_shape = (window_size, nb_features)
model = autoencoderLstm(input_shape)
data_files, add_to_name, data_config = common_code_normalized()
result_files = data_files
result_files = use_whole_data(data_files, input_shape,
train_autoencoder_based_models, model)
write_result(algorithm_name='autoencoderLstm',
data_files=result_files,
results_path=cwd + '/results')
algo_name = "autoencoderLstmOneEpoch{}{}{}{}".format(now.month, now.day,
now.hour, now.minute)
with open("dump/" + algo_name + ".obj", 'wb') as f:
pickle.dump(result_files, f)
write_result(algorithm_name=algo_name,
data_files=result_files,
results_path=cwd + '/results')
def print_TP(self):
utility.write_result('true_positive.result', list(self.TP))
def print_FP(self):
print "False Positive pairs: %d" % len(self.FP)
utility.write_result('false_positive.result', list(self.FP))
def detect_type3(self):
self.type.detector_type3()
type3_result = self.type.similar_pairs
utility.write_result("./type3.dup", type3_result)