Table 3.

The performances of the AA, MDS, and AML three-classification model with different outcome weights.

Models	Data set	Sensitivity (95%CI)	Specificity (95%CI)	PPV (95%CI)	NPV (95%CI)	AUC (95%CI)	Accuracy (95%CI)
30 epochs, 5:5 outcome weight	Training set	0.884 (0.867-0.902)	0.960 (0.952-0.968)	0.922 (0.907-0.937)	0.940 (0.930-0.949)	0.970 (0.965-0.976)	0.934 (0.926-0.942)
	Testing set	0.834 (0.803-0.865)	0.952 (0.939-0.965)	0.902 (0.876-0.928)	0.915 (0.898-0.931)	0.945 (0.934-0.957)	0.911 (0.897-0.925)
	Validate set	0.858 (0.826-0.891)	0.880 (0.858-0.901)	0.787 (0.751-0.823)	0.923 (0.905-0.941)	0.911 (0.892-0.929)	0.872 (0.854-0.890)
30 epochs, 2:8 outcome weight	Training set	0.855 (0.836-0.874)	0.952 (0.943-0.960)	0.905 (0.888-0.921)	0.925 (0.914-0.935)	0.971 (0.966-0.976)	0.918 (0.909-0.927)
	Testing set	0.807 (0.774-0.839)	0.929 (0.913-0.944)	0.858 (0.828-0.888)	0.900 (0.882-0.918)	0.945 (0.933-0.956)	0.886 (0.870-0.902)
	Validate set	0.823 (0.788-0.858)	0.889 (0.868-0.910)	0.793 (0.757-0.830)	0.906 (0.887-0.926)	0.905 (0.886-0.924)	0.866 (0.848-0.885)
30 epochs, 1:9 outcome weight	Training set	0.890 (0.873-0.907)	0.986 (0.981-0.990)	0.970 (0.961-0.980)	0.944 (0.935-0.953)	0.976 (0.971-0.981)	0.952 (0.945-0.959)
	Testing set	0.841 (0.810-0.871)	0.972 (0.962-0.982)	0.941 (0.921-0.962)	0.920 (0.903-0.936)	0.958 (0.948-0.968)	0.926 (0.913-0.939)
	Validate set	0.852 (0.819-0.885)	0.901 (0.882-0.921)	0.817 (0.783-0.852)	0.921 (0.903-0.940)	0.925 (0.909-0.941)	0.884 (0.867-0.902)

CI, confidence interval; PPV, positive predictive value; NPV, negative predictive value; AUC, area under the curve.