Table 2 Comparison between STL and MTL in diabetes mellitus prediction. MAND-LR, MAND-MLP, MAND-LSTM, and MAND-MHSA denote the MAND architecture integrated with logistic regression, multilayer perceptron (MLP), LSTM, and multi-head self-attention as ICD feature extraction modules, respectively. FM and DCN represent CTR-based approaches. BAC: balanced accuracy; FPR: false positive rate; FNR: false negative rate.

Backbone model	STL/MTL	Log loss	AUC	BAC	Precision	Recall	F1 score	FPR	FNR
MAND-LR¹⁸	STL	0.3043	0.8818	0.7588	0.9331	0.5284	0.6748	0.0108	0.4716
MAND-LR¹⁸	MTL	0.3084	0.8803	0.7563	0.9337	0.5233	0.6707	0.0107	0.4767
MAND-MLP¹⁸	STL	0.2906	0.8852	0.7665	0.9638	0.5387	0.6911	0.0057	0.4613
MAND-MLP¹⁸	MTL	0.2967	0.8831	0.7607	0.9598	0.5277	0.6810	0.0063	0.4723
MAND-LSTM¹⁸	STL	0.2858	0.8926	0.7791	0.9436	0.5680	0.7091	0.0098	0.4320
MAND-LSTM¹⁸	MTL	0.2850	0.8912	0.7728	0.9543	0.5533	0.7005	0.0077	0.4467
MAND-MHSA¹⁸	STL	0.2888	0.8918	0.7822	0.9003	0.5829	0.7076	0.0185	0.4171
MAND-MHSA¹⁸	MTL	0.2924	0.8900	0.7760	0.9019	0.5698	0.6984	0.0178	0.4302
FM¹⁹	STL	0.3378	0.8749	0.7635	0.8557	0.5538	0.6725	0.0268	0.4462
FM¹⁹	MTL	0.3250	0.8699	0.7551	0.8761	0.5319	0.6620	0.0217	0.4681
DCN²²	STL	0.2871	0.8899	0.7749	0.9493	0.5584	0.7031	0.0086	0.4416
DCN²²	MTL	0.2914	0.8870	0.7690	0.9495	0.5464	0.6936	0.0084	0.4536

Quick links

Search