Monitoring stance towards vaccination in twitter messages

Table 5 Machine Learning performance of correctly predicting the label of tweets with a negative stance (Clf = Classifier, NB = Naive Bayes, SVM = Support Vector Machines, AUC = Area under the curve)

		Binary		Irrelevance filter		Polarity		Polarity - Sentiment
Training data	Clf	F1	AUC	F1	AUC	F1	AUC	F1	AUC
Strict	NB	0.14	0.53	0.15	0.54	0.24	0.56	0.30	0.60
	SVM	0.30	0.59	0.32	0.61	0.34	0.62	0.35	0.63
Strict + Lax	NB	0.26	0.58	0.27	0.59	0.33	0.63	0.36	0.64
	SVM	0.33	0.63	0.34	0.63	0.36	0.66	0.36	0.64
Strict + One	NB	0.13	0.53	0.15	0.54	0.24	0.57	0.27	0.59
	SVM	0.29	0.59	0.29	0.59	0.34	0.62	0.37	0.64
Strict + Lax + One	NB	0.27	0.58	0.27	0.59	0.33	0.62	0.32	0.61
	SVM	0.34	0.63	0.32	0.62	0.35	0.64	0.36	0.64

ISSN: 1472-6947