Table 2.

Performance of the machine learning based methods on various transfer settings.

Source task(s) (training data)	Target task (test data)	Method	P†	R†	F1†
HOST-LEVEL TRANSFER
Salmonella-human	Salmonella-mouse	Baseline	42.8	93.7	58.8
		T-SVM	45.4	93.7	61.2
		KMM-SVM	51.7	93.7	66.7
Salmonella-mouse	Salmonella-human	Baseline	95.4	33.8	50
		T-SVM	67.5	43.5	52.9
		KMM-SVM	100	35.5	52
PATHOGEN-LEVEL TRANSFER
Francisella-human, E.coli-human	Salmonella-human	Baseline	17.8	12.9	14.9
		T-SVM	15	14.5	14.7
		KMM-SVM	25.7	16.1	19.9
Francisella-human, Salmonella-human	E.coli-human	Baseline	12.9	12.5	12.7
		T-SVM	10.4	15.6	12.5
		KMM-SVM	15.9	21.9	18.4

We compare them with a simple baseline: inductive kernel-SVM. We report precision (P), recall (R) and f-score (F1). The data that was used to build each of the models is shown in the first column. The second column shows the target task—the data on which we evaluate the model. The numbers in bold font indicate the highest performance in that column (i.e., for that metric).

^†Computed using the default classifier threshold: 0.5.

The positive:negative class ratio in all datasets was 1:100.

The performance of a random classifier would be F-score = 1.