Table 1.

Depression classification performance of a support vector machine (SVM) and random forest (RF) model with varying sample size.

Model	AUC	Accuracy	F1	Sensitivity	Specificity
SVM (n = 1,006)	0.56	0.59	0.42	0.33	0.79
SVM (n = 476)	0.59	0.59	0.54	0.52	0.66
RF (n = 1,006)	0.56	0.58	0.44	0.38	0.75
RF (n = 476)	0.57	0.57	0.53	0.52	0.63
De Choudhury SVM (n = 476)	NA	0.68	NA	0.58	NA

Depression classification performance was similar between SVM and RF regardless of sample size. In the reduced sample size (n = 476) of participants with high word counts there was an increase in the F1 score and sensitivity for both the SVM and RF models. However, in both models the increase in sensitivity was accompanied by a reduction in specificity. Neither model improved over the classification performance of de Choudhury et al.⁵.