Analysis and Recognition of Food Safety Problems in Online Ordering Based on Reviews Text Mining

Table 4

Evaluation results of confusion matrix of text distinguish classifier of the online ordering platform.

Classification algorithm	Accuracy				Precision				Specificity				F1 measure
	BERT	TF-IDF	Word2vec	N-gram	BERT	TF-IDF	Word2vec	N-gram	BERT	TF-IDF	Word2vec	N-gram	BERT	TF-IDF	Word2vec	N-gram
GBDT	0.908	0.899	0.880	0.899	0.930	0.901	0.887	0.901	0.531	0.296	0.187	0.296	0.936	0.943	0.930	0.943
XGBoost	0.891	0.892	0.851	0.871	0.911	0.892	0.852	0.871	0.388	0.218	0.030	0.177	0.928	0.941	0.918	0.929
LSTM	0.781	0.803	0.871	0.853	0.869	0.803	0.877	0.853	0.136	0.000	0.099	0	0.828	0.891	0.927	0.920
BiLSTM	0.779	0.866	0.860	0.853	0.867	0.866	0.861	0.853	0.126	0.000	0.007	0	0.827	0.928	0.924	0.921
CNN	0.809	0.813	0.801	0.863	0.865	0.870	0.866	0.863	0.068	0.105	0.088	0	0.862	0.864	0.853	0.927
RNN	0.834	0.853	0.850	0.861	0.867	0.853	0.850	0.861	0.054	0.000	0.000	0	0.890	0.921	0.919	0.925
RCNN	0.804	0.798	0.815	0.814	0.866	0.865	0.864	0.870	0.085	0.085	0.048	0.109	0.855	0.850	0.870	0.865