Automated Fraudulent Phone Call Recognition through Deep Learning

Table 1

Tuned hyperparameters of the selected DL models.

Hyperparameter	CNN	LSTM	SDAE
Optimizer	RMSProp	Adam	RMSProp
Batch size	512	512	512
Training epochs	10	300	500
Number of layers	8	2	5
Input units	176	176	176
Dropout	0.25	0.2	0.1
Activation	ReLU	Sigmoid	Sigmoid
Kernels	64/128	—	—
Kernel size	3	—	—
Pool size	2	—	—