BUPNN: Manifold Learning Regularizer-Based Blood Usage Prediction Neural Network for Blood Centers

Table 3

Classification AUC comparison with the baseline methods, the best result is shown in bold. The second result is italicized. The brackets at the right end show how much BUPNN exceeds the optimal metrics in the other methods.

	KNN	RF	MLP	ET	SVM	GB	AdaB	LGBM	BUPNN
NC-A	0.8072	0.9166	0.8436	0.8443	0.8736	0.8949	0.9072	0.8881	0.9229 (0.0063)
NC-B	0.7302	0.8109	0.7421	0.8178	0.7116	0.7214	0.8324	0.8119	0.8349 (0.0240)
COM-mid	0.8009	0.8526	0.7764	0.8437	0.7435	0.8442	0.842	0.8508	0.8843 (0.0317)
COM-mea	0.8054	0.8591	0.8399	0.8252	0.7553	0.8470	0.8562	0.8630	0.8797 (0.0167)
COM-KNN	0.8033	0.8575	0.7912	0.8321	0.7739	0.8446	0.8526	0.8620	0.8761 (0.0141)
Average	0.7894	0.8593	0.7992	0.8326	0.7716	0.8304	0.8581	0.8552	0.8796 (0.0202)