Investigating the impact of pre-processing techniques and pre-trained word embeddings in detecting Arabic health information on social media

Journal of Big Data

Table 13 Optimal hyperparameters of the CNN

	CNN_Mazajak sg	CNN_Mazajak cbow	CNN_fasttext	CNN_arwordvec sg	CNN_arwordvec cbow	Cnn arvec sg	Cnn arvec cbow
cov_filter	32	32	1	1	32	1	32
cov_filter1	32	32	1	32	32	32	32
cov_filter2	1	32	1	32	32	1	32
cov_kernel	32	32	32	32	32	1	1
cov_kernel1	1	1	32	1	32	32	32
cov_kernel2	1	1	1	1	32	1	1
pool_filter	32	32	1	32	1	1	1
cov1_activation	relu	sigmoid	relu	relu	relu	sigmoid	relu
cov1_activation1	relu	relu	relu	relu	sigmoid	relu	relu
cov1_activation2	sigmoid	relu	relu	relu	relu	relu	relu
dropout_1	0.0	0.0	0.6	0.600	0.600	0.300	0.000
dense_units	380	20	20	20.000	380.000	20.000	380.000
Dense activatino	relu	relu	sigmoid	relu	relu	sigmoid	sigmoid
dropout_2': 0.0	0.0	0.000	0.5	0.0	0.0	0.5	0.0
learning_rate	0.01	0.001	0.001	0.001	0.001	0.001	0.001
Batch size	40	40	40	40	40	40	300