An Assessment of Lexical, Network, and Content-Based Features for Detecting Malicious URLs Using Machine Learning and Deep Learning Models

Table 4

Parameters’ settings applied for all classifiers.

Model	Parameter	Value
RF	Number of trees	100
CNN	Activation function in hidden layers	ReLU
	Number of neurons in output layer	1
	Activation function in output layer	Sigmoid
	Dropout	0.2, 0.5
	Batch size	32
	Number of layers	4
	Number of neurons in hidden layers	32, 64, 64
LSTM	Activation function in hidden layers	Tanh
	Number of neurons in output layer	1
	Activation function in output layer	Sigmoid
	Dropout	0.1
	Batch size	32
	Number of layers	3
	Number of neurons in hidden layers	8, 8