Machine learning to refine decision making within a syndromic surveillance service

BMC Public Health

Table 4 Performance measures (tenfold cross validation) for multi-state classification of statistical alarms in the PHE multi-system syndromic surveillance service

			NB	TAN	TAN*	Multinet	Multinet*
Decision	Alert	MCC	0.398	0.377	0.393	0.387	0.435
		Precision	0.308	0.420	0.355	0.846	0.490
		Recall	0.516	0.339	0.435	0.177	0.387
	Monitor	MCC	0.497	0.581	0.552	0.486	0.459
		Precision	0.354	0.570	0.450	0.524	0.259
		Recall	0.708	0.597	0.684	0.455	0.829
	Alert + Monitor (merging decisions/classifications from Table 3)	MCC	0.587	0.643	0.617	0.521	0.507
		Precision	0.422	0.641	0.510	0.595	0.303
		Recall	0.826	0.649	0.754	0.462	0.867

*Modified approach to account for data asymmetry i.e the predominance of ‘No-Action’ outcomes

ISSN: 1471-2458