TABLE 3.

Performance of PREFUM, comparing predicted total fumonisin classes versus observed total fumonisin classes.

		Predictiona
		Training set			Internal validation set			External validation set
		High	Low	Accuracy	High	Low	Accuracy	High	Low	Accuracy
Observationb	High	28	11	73%	10	6	76%	4	4	80%
	Low	18	51		3	18		5	32

^aThe fumonisin contamination levels of samples from different datasets predicted by the model. ^bThe measured fumonisin contamination levels of samples from different datasets.