LA-GRU: Building Combined Intrusion Detection Model Based on Imbalanced Learning and Gated Recurrent Unit Neural Network

Table 3

Size and distribution of samples in new dataset.

Algorithm	Parameter	Value
LA-SMOTE	Number of nearest neighbors	65
	Over-sampling rate	600%
GRU	Number of nodes in input layer	122
	Number of neurons in hidden layer	75
	Number of neurons in output layer	5
	Batch size	500
Adam	Step size	0.001
	First-order exponential damping decrement	0.9
	Second-order exponential damping decrement	0.999
	Non-zero constant	10^−8