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

Table 6

MSE comparison for all data, best results are shown in bold; results with clear advantage are shown in underline. The second result is italicised. The brackets at the right end show how much BUPNN exceeds the optimal metrics in the other methods.

	LR	SVM	MLP	ET	GB	RF	LGBM	Adab	BUPNN
NC-A	0.0095	0.0112	0.0121	0.0194	0.0093	0.0090	0.0090	0.0161	0.0079
COM-mid	0.0042	0.0075	0.0167	0.0103	0.0040	0.0043	0.0042	0.0175	0.0038
COM-mea	0.0042	0.0076	0.0076	0.0074	0.0039	0.0041	0.0044	0.0151	0.0037
COM-KNN	0.0041	0.0065	0.0082	0.0142	0.0040	0.0039	0.0043	0.0109	0.0036
Average	0.0055	0.0082	0.0111	0.0128	0.0053	0.0053	0.0054	0.0149	0.0048