Monitoring stance towards vaccination in twitter messages

Table 6 Performance of all systems on correctly predicting the label of tweets with a negative stance (for ML only the best performing system is displayed; Pr = Precision, Re = Recall, AUC = Area under the Curve)

	Pr	Re	F1	AUC
Random (50%)	0.11	0.46	0.18	0.48
Random (15%)	0.12	0.15	0.13	0.50
Pattern	0.14	0.34	0.20	0.53
Coosto	0.20	0.31	0.25	0.57
Best ML system	0.29	0.47	0.36	0.66
Ensemble system (precision optimized)	0.34	0.19	0.25	0.57
Ensemble system (recall optimized)	0.18	0.61	0.28	0.62

ISSN: 1472-6947