A Multimodel-Based Deep Learning Framework for Short Text Multiclass Classification with the Imbalanced and Extremely Small Data Set

Table 1

Method comparison.

Model	Precision (%)	Recall (%)	F1-score (%)	Time (min)	Total params	Trainable params	Epoch	Size (MB)
BERT + BI-GRU	60	54	54	241	109,920,145	1,609,873	6	419
BERT-BI-LSTM	61	53	54	380	109,247,003	936,731	11	417
BERT + CNN	58	51	52	153	108,404,591	94,319	4	413
BERT–CNN–LSTM	57	51	52	301	109,200,743	890,471	8	416
BERT-LSTM-CNN	52	44	44	189	111,918,864	3,608,592	5	427
DistilBERT + BI-GRU	59	54	54	85	66,800,785	1,609,873	5	261
DistilBERT BI-LSTM	60	54	54	84	66,127,643	936,731	5	252
DistilBERT-CNN	56	53	53	69	65,285,231	94,319	4	249
DistilBERT–CNN–LSTM	59	50	51	132	66,081,383	890,471	7	252
DistilBERT-LSTM-CNN	56	46	48	75	68,799,504	3,608,592	4	268