Table 3.

The tuned RF1 machine learning model confusion matrix for the assignment of 322 training set animal origin S. Typhimurium and monophasic S. Typhimurium isolates to eight primary source classes.

	Broilers	Cattle	Game	Layers	OtherMammals	Pigs	Sheep	Turkey
Broilers	15	1	0	0	0	0	0	0
Cattle	0	60	0	0	7	1	8	0
Game	0	0	14	0	0	0	0	0
Layers	0	0	0	5	0	0	0	0
OtherMammals	0	0	0	0	36	0	1	0
Pigs	1	0	0	1	2	131	0	0
Sheep	0	1	0	0	0	0	29	0
Turkey	0	0	0	0	0	0	0	9
Sensitivity	0.938	0.968	1.000	0.833	0.800	0.992	0.763	1.000
Specificity	0.997	0.939	1.000	1.000	0.996	0.979	0.996	1.000
Balanced accuracy	0.967	0.953	1.000	0.917	0.898	0.986	0.880	1.000

The values along the diagonal (in bold) indicate the number of isolates correctly assigned by the model to their actual primary source class. The values above and below the diagonal indicate the number of isolates incorrectly classed by the model not to their actual primary source class (column headers) but to the model predicted source (row names). The isolates were assigned to the primary source class with the highest model computed probability of assignment. Balanced accuracy is the average of the sensitivity (true positive rate) and specificity (true negative rate) values for each primary source class.