Phase 2 single-mode and dual-mode classification. (a)
Salmon fillet
classification accuracies ±1 day for 5-fold cross validation
and test data sets for single-mode and dual-mode fusion for fluorescence
(Flu), visible absorbance (vis), and NIR absorbance (NIR) phenomena,
for four different classification algorithms: LDA, WKNN, Gaussian
naïve (GN), and EBT. Accuracies color-coded: green: 90–100%;
yellow: 80–89%; orange: 70–79%; red: 60–70%;
brown 50–60%; and gray: no classification. “# of PCs”
refers to the number of PCs retained for each classification model.
An optimal number of PCs were selected for each model. Data were acquired
from a second salmon fillet/set of experiments (Fillet 2, Phase 2).
Data graphed in (b) for (i) single-mode and (ii) dual-mode. CV and
test data are plotted for each algorithm (color in print/online).