Table 2.

Overall performances of RFs, gpt-3.5-turbo, and BERT models.

		Random forests			gpt-3.5-turbo			BERT
Models	Categories	P	R	F	P	R	F	P	R	F
One-step (end-to-end)	Diet	0.910	0.919	0.915	0.793	0.970	0.873	0.981	0.966	0.973
	Exercise	0.946	0.906	0.926	0.711	1.000	0.831	0.953	0.979	0.965
	Mental health	0.916	0.784	0.844	0.485	0.979	0.648	0.951	0.943	0.945
	Health literacy	0.914	0.740	0.818	0.867	0.980	0.920	0.932	0.914	0.921
	Unrelated	0.902	0.977	0.938	1.000	0.599	0.749	0.970	0.954	0.962
	Macro	0.917	0.865	0.888	0.771	0.906	0.804	0.957	0.951	0.953
Two-step	Diet	0.876	0.928	0.899	0.932	0.970	0.951	0.918	0.971	0.943
	Exercise	0.890	0.938	0.909	0.795	0.969	0.873	0.910	0.953	0.927
	Mental health	0.832	0.889	0.856	0.688	0.979	0.808	0.888	0.897	0.892
	Health literacy	0.884	0.820	0.849	1.000	0.840	0.913	0.923	0.873	0.895
	Unrelated	0.957	0.937	0.947	0.966	0.854	0.906	0.957	0.937	0.947
	Macro	0.888	0.902	0.892	0.876	0.922	0.890	0.919	0.926	0.921

P, Precision; R, Recall; F, F1-score.

The macro-averaged values are italicized and the highest F1-scores are underlined.