Table 2.

Results of random forest and artificial neural network classification models to predict the multi-nutritional classes of black (variable purple and purple) rice. Abbreviated variables: L- whiteness, A- redness, B- yellowness, TAC – total anthocyanin content, TFC – total flavonoid content, TPC – total phenolic content, FC – flavonoid content, PC – phenolic content, AC – anthocyanin content.

Model	Predictors	Number of Nodes in Hidden Layer	Model Accuracy
Artificial Neural Network (2019 Dry Season, n = 327)	L, A, B, Hue, Intensity	8	36.42%
	Bound PC, TAC, Mo, P, Bound FC, Fe, Ca, TPC, TFC, K, Mn, L, Al, Zn, Cu, Na. A. B. Bound AC	23	85.35%
	Bound PC, TAC, Mo, P, Bound FC, L, A, B	9	99.9%
Random Forest (2019 Dry Season, n = 327)	L, A, B, Hue, Intensity	Not Applicable	30.21%
	Bound PC, TAC, Mo, P, Bound FC, Fe, Ca, TPC, TFC, K, Mn, L, Al, Zn, Cu, Na. A. B. Bound AC	Not Applicable	78.47%
	Bound PC, TAC, Mo, P, Bound FC, L, A, B	Not Applicable	85.3%
Artificial Neural Network (2020 Wet Season Validation Set, n = 200)	Bound PC, TAC, Mo, P, Bound FC, L, A, B	9	87.6%
Random Forest (2020 Wet Season Validation Set, n = 200)	Bound PC, TAC, Mo, P, Bound FC, L, A, B	Not Applicable	75.43%