Weibo Rumor Recognition Based on Communication and Stacking Ensemble Learning

Table 4

The effectiveness of each feature in .

Ranking	Removed	Class	Accuracy	Precision	Recall	F1-score
1	The number of friends	N	0.9344	0.9573	0.9038	0.9298
		R		0.9159	0.9630	0.9388
2	The number of #	N	0.9344	0.9552	0.9060	0.9300
		R		0.9175	0.9609	0.9387
3	Emotional Consistency	N	0.9343	0.9531	0.9083	0.9301
		R		0.9191	0.9588	0.9386
4	The number of mutual followers	N	0.9335	0.9577	0.9028	0.9294
		R		0.9131	0.9630	0.9373
5	The number of followers	N	0.9334	0.9551	0.9038	0.9287
		R		0.9157	0.9609	0.9378
6	The number of @	N	0.9334	0.9487	0.9105	0.9292
		R		0.9206	0.9547	0.9374
7	Picture	N	0.9334	0.9529	0.9060	0.9289
		R		0.9173	0.9588	0.9376
8	Verified	N	0.9332	0.9508	0.9083	0.9291
		R		0.9190	0.9568	0.9375
9	User Credibility	N	0.9332	0.9508	0.9083	0.9291
		R		0.9190	0.9568	0.9375
10	The number of microblogs	N	0.9331	0.9541	0.9060	0.9294
		R		0.9180	0.9612	0.9391
11	The length of text	N	0.9325	0.9507	0.9060	0.9278
		R		0.9172	0.9568	0.9366
12	User Influence	N	0.9325	0.9550	0.9016	0.9275
		R		0.9139	0.9609	0.9368
13	The number of reports	N	0.9303	0.9505	0.9016	0.9254
		R		0.9136	0.9568	0.9347
14	The number of likes	N	0.9303	0.9401	0.9128	0.9262
		R		0.9218	0.9465	0.9340
15	Time span	N	0.9293	0.9441	0.9060	0.9247
		R		0.9167	0.9506	0.9333
16	Regional Correlation	N	0.9175	0.9405	0.8837	0.9112
		R		0.8986	0.9486	0.9229