Predicting subjective well-being in a high-risk sample of Russian mental health app users

EPJ Data Science

Table 9 Best WHO-5 classification results

Classifi cation	Thre-shold	N (Classes)	Best model	Best features	F1-macro	F1-weigh-ted	F1-low	F1-high	True Positive Rate (low)	False Positive Rate (low)
Binary	0.51	221/151	Ada-Boost	Words + RuLIWC + AppCats	0.692	0.706	0.768	0.616	0.792	0.404
Binary majority baseline					0.378	0.456	0.373	0	1.0	1.0
Trinary	0.35/0.59	111/158/103	Ada-Boost	Clusters + RuLIWC + Words	0.483	0.493	0.502	0.433	0.450	0.161
Trinary majority baseline					0.199	0.253	–	–	0.0	0.0