Table 4.

Holdout mean squared error (MSE) for continuous models trained on oversampled training data using linear, random forest, deep neural network (DNN) and ordinal regression, and random forest and DNN split classification.

Model	Continuous MSE	Ordinal MSE	Cohen’s Kappa
Linear regression	0.373	0.459	0.351
Random forest regression	0.337	0.437	0.412
DNN regression	0.374	0.471	0.420
Ordinal regression	0.373	0.458	0.352
Random forest split classifiers	0.387	0.440	0.456
DNN split classifiers	0.683	0.739	0.413