The Journal of Headache and Pain

Table 4 Experimental results for multi-class classification and cluster headache class detection

From: Using natural language processing to automatically classify written self-reported narratives by patients with migraine or cluster headache

Two Classes
	Naive Bayes Classifier				Support Vector Machine					Logistic Regression
	P	R	F1-score	Accuracy	P	R	F1-score		Accuracy	P	R	F1-score	Accuracy
N-grams	0,744	0,7	0,688	0,732	0,854	0,832	0,838		0,858	0,856	0,83	0,838	0,858
Metadata	0,802	0,816	0,808	0,821	0,804	0,816	0,808		0,821	0,804	0,816	0,808	0,821
N-grams + metadata	0,744	0,7	0,688	0,732	0,838	0,826	0,828		0,849	0,848	0,84	0,84	0,858
Class: cluster headache
	Naive Bayes Classifier			Support Vector Machine				Logistic Regression
	P	R	F1-score	P	R	F1-score		P	R	F1-score
N-grams	0,676	0,586	0,586	0,834	0,74	0,778		0,838	0,74	0,779
Metadata	0,71	0,81	0,754	0,71	0,81	0,754		0,71	0,81	0,754
N-grams + metadata	0,676	0,586	0,586	0,792	0,76	0,769		0,808	0,784	0,79

Legend: Highest accuracy scores for the two classes are boldfaced, the best F1-score for the ‘cluster headache’ class is underlined. Abbreviations: Avg Average, P Precision, R Recall

Back to article page

ISSN: 1129-2377

Contact us

Submission enquiries: journalsubmissions@springernature.com