表 2. Comparison results between the method in this paper and the single-task classification method.

本文方法与单任务分类方法的对比结果

算法	ACC	AUC	REC	PPV	NPV	F1
He等[18]	78.44% ± 3.36%	79.49% ± 4.73%	66.19% ± 14.75%	69.90% ± 3.73%	77.19% ± 4.16%	63.45% ± 8.32%
Huang等[19]	79.11% ± 4.23%	77.25% ± 4.51%	63.84% ± 8.15%	64.50% ± 7.10%	73.01% ± 5.96%	59.80% ± 8.04%
Carreira等[20]	78.89% ± 1.92%	81.06% ± 3.86%	71.05% ± 12.41%	68.79% ± 9.31%	81.10% ± 3.13%	65.34% ± 8.02%
Zhou等[9]	78.22% ± 1.68%	79.25% ± 1.43%	67.67% ± 3.95%	68.32% ± 6.35%	77.86% ± 2.90%	63.30% ± 3.70%
Yang等[3]	80.67% ± 1.15%	80.47% ± 1.06%	60.89% ± 3.30%	68.50% ± 6.73%	78.24% ± 2.01%	61.22% ± 2.23%
本文算法	80.67% ± 0.67%	79.09% ± 3.16%	86.08% ± 0.83%	80.07% ± 1.95%	62.67% ± 2.54%	80.05% ± 1.70%