Combining the External Medical Knowledge Graph Embedding to Improve the Performance of Syndrome Differentiation Model

Table 2

The results on the EMRs dataset.

Models	P@1	P@3	P@5	Average precision	Hamming loss	Label ranking loss
AttentiveConvNet	0.416	0.404	0.384	0.263	0.081	0.136
DPCNN	0.690	0.63	0.574	0.528	0.045	0.079
Transformer	0.721	0.665	0.61	0.578	0.04	0.086
AttentionXML	0.749	0.702	0.645	0.537	0.043	0.078
TextCNN	0.753	0.686	0.636	0.603	0.039	0.069
TextRNN	0.769	0.719	0.668	0.646	0.034	0.068
FastText	0.809	0.748	0.686	0.676	0.034	0.046
BERT	0.89	0.876	0.845	0.857	0.026	0.016
KG-based	0.059	0.073	0.066	—	—	—
KDSD	0.9	0.879	0.852	0.866	0.029	0.014