Table 3.

Average model performance result by 10-fold cross-validation in each drag-and-drop subgame.

Deep learning model for individual subgame	Performance, mean (SD)
	AUROCa	Accuracy	F score	Precision	Recall	Specificity
Second subgame	0.746 (0.116)	0.719 (0.094)	0.627 (0.165)	0.616 (0.136)	0.683 (0.151)	0.755 (0.089)
Third subgame	0.793 (0.117)	0.747 (0.082)	0.675 (0.126)	0.686 (0.142)	0.688 (0.167)	0.783 (0.069)
Sixth subgame	0.817 (0.070)	0.769 (0.078)	0.708 (0.153)	0.675 (0.183)	0.757 (0.123)	0.740 (0.099)

^aAUROC: area under the receiver operating characteristics curve.