Table 2.

Hyperparameter setting.

	Hyperparameter	Value
BiLSTM	Number of layers	1
	Hidden size	32
	Dropout	0.05
	Activation functions	Tanh
TCN	TCN layers	2
	Kernel size	3
	Dropout	0.1
	Activation functions	Relu
DCEFormer	Sensor encoder blocks	2
	Sensor encoder heads	4
	Timestep encoder blocks	2
	Timestep encoder heads	4
	Decoder blocks	1
	Decoder heads	4
	Activation functions	Softmax
Other	Adam learning rate	0.0005
	Batch size	32
	Fully connected layer activation functions	Relu
	Dropout	0.05
	Early stopping patience	10