Deep Learning Structure for Cross-Domain Sentiment Classification Based on Improved Cross Entropy and Weight

Table 8

Experimental results under different word vector dimensions.

	Word vector dimension	50	100	150	200	250
LSTM-BO	Accuracy (%)	82.14	82.48	81.69	81.76	82.04
	Loss (%)	0.2267	0.2234	0.2372	0.2577	0.2302
CNN-BO	Accuracy (%)	88.74	87.66	88.56	88.34	87.46
	Loss (%)	0.1637	0.1845	0.1652	0.1732	0.1832