Table 2.

Performance metrics for multi-class diagnosis prediction

Model	G	Sensitivity
Reinforcement learning (Q-imb)	0.834 (0.082)	0.748 (0.126)
Reinforcement learning (DDQN)	0.776 (0.096)	0.671 (0.153)
Reinforcement learning (DQN) Lin et al. (2020)	0.777 (0.107)	0.672 (0.172)
Neural network	0.806 (0.105)	0.715 (0.183)
Neural network + SMOTE	0.804 (0.109)	0.714 (0.193)
Neural network + Cost-Sensitive	0.801 (0.103)	0.712 (0.190)
XGBoost	0.819 (0.106)	0.733 (0.181)
XGBoost + SMOTE	0.819 (0.107)	0.733 (0.184)
XGBoost + Cost-sensitive	0.830 (0.092)	0.744 (0.142)

Results reported as mean sensitivities and G values across all classes, shown alongside standard deviation

Bold and italics values denote best and second best scores, respectively, for G-mean