Fig 1.
± 1.96 standard deviations) across all ML approaches with no reduction (NR), PCA and fPCA using n-fold and farm-fold validation methods.
Fig 2.
Multiple model performance metrics (mean
± 1.96 standard deviations) for random forest applied to the accelerometer data under the three approaches (NR, PCA, fPCA) and under the two cross validation methods (n-fold and farm-fold).
Fig 3.
Calibration plots for random forest applied to accelerometer data under three approaches (NR, PCA, fFPCA) and two cross validation methods (n-fold and farm-fold).
Dots represent mean per bin with vertical lines indicating 95% confidence intervals. Numbers represent number of animals per bin. Twenty bins were used and those with <5 predicted observations are not plotted.