An End-to-End Rumor Detection Model Based on Feature Aggregation

Table 7

Early detection of rumor (R: rumor; N: nonrumor).

Deadline (h)	Label	Accuracy	Precision	Recall
1	R	0.908	0.878	0.952	0.913
	N		0.945	0.862	0.901
2	R	0.928	0.920	0.941	0.931
	N		0.937	0.914	0.926
6	R	0.912	0.930	0.895	0.912
	N		0.895	0.930	0.912
12	R	0.924	0.913	0.941	0.927
	N		0.937	0.906	0.921
24	R	0.941	0.949	0.935	0.942
	N		0.933	0.947	0.940
36	R	0.944	0.936	0.956	0.946
	N		0.953	0.932	0.942
48	R	0.927	0.944	0.912	0.928
	N		0.911	0.943	0.927
72	R	0.942	0.940	0.948	0.944
	N		0.945	0.936	0.941
96	R	0.940	0.924	0.962	0.943
	N		0.959	0.917	0.937