Algorithm 1: One-Class Loss

Input: Training set, # Normal data             Evaluating set, # Normal data and Unnormal data = 1:1 Output: Normal label, Unnormal label Process: #step 1: Bi-GRU AE train          model = Bi-GRU AE) # init model          model.fit(TrainData, split = 0.2, batchsize, epoch)         PredictData = model.predict(TrainData)         Loss = abs(TrainData-PredictData)#absolute value         Loss = sort(Loss)         Loss_train = max(Loss)         Return Loss_train # step 2: One-Class Classification         PredictData = model.preict(TestData)         Loss = abs(TrainData-PredictData)         Loss = sort(Loss))         Loss = max(Loss)         If Loss > Loss_train:            Output: Unnormal label         Else:            Output: Normal label