Assessing the Influence Level of Food Safety Public Opinion with Unbalanced Samples Using Ensemble Machine Learning

Table 3

Influence assessment results.

Model		Evaluation index
		Accuracy	Precision	Recall	F1
NB	Original sample	0.7587	0.7615	0.7587	0.7592
	Balanced sample	0.8126	0.8127	0.8126	0.8124
SVM	Original sample	0.7968	0.7990	0.7968	0.7972
	Balanced sample	0.8435	0.8439	0.8435	0.8434
XGBoost	Original sample	0.7902	0.7918	0.7902	0.7905
	Balanced sample	0.8491	0.8490	0.8491	0.8490
FastText	Original sample	0.8027	0.8037	0.8027	0.8030
	Balanced sample	0.8470	0.8468	0.8469	0.8469
TextCNN	Original sample	0.7870	0.7907	0.7870	0.7879
	Balanced sample	0.8392	0.8417	0.8392	0.8393
LSTM	Original sample	0.8018	0.8064	0.8018	0.8024
	Balanced sample	0.8494	0.8513	0.8494	0.8494
BERT + summarization	Original sample	0.7919	0.7934	0.7919	0.7921
	Balanced sample	0.8300	0.8298	0.8300	0.8299
Ensemble model (ours)	Original sample	0.8062	0.8071	0.8062	0.8052
	Balanced sample	0.8530	0.8541	0.8530	0.8534