Text Sentiment Analysis Based on a New Hybrid Network Model

Table 2

Training parameter settings of doc2vec + CNN + BiLSTM + Attention model.

Name	Parameter	Parameter value
doc2vec embedded layer	Embeddings_dim	200
	Max_len	200
Convolution layer	Filters	64
	Kernel_size	5
	Padding	Valid
	Activation	Relu
	Strides	1
Convolution layer	Pool_size	4
BiLSTM model	Units	128
	Dropout	0.5
Full connection layer	Units	64
	Activation	Relu
Output layer	Units	1
	Activation	Sigmoid
Training parameters	Optimizer	Adam
	Loss	Binary_cross-entropy
	Metrics	Accuracy
	Rate	0.01
	Batch_size	32
	Epoch	5