An Improved Multitask Learning Model with Matching Network and Its Application in Traditional Chinese Medicine Syndrome Recommendation

Table 7

Match-net extract features compared with designed features.

Model	Hits@10	MeanRank	MRR	AUC
MMOE	0.76824	2.98508	0.33500	0.90738
Only match-net	0.75623	2.99645	0.33372	0.89786
Zxd1:avg|sum	0.77932	2.90490	0.34426	0.90702
tfidf:avg|sum	0.77980	2.90894	0.34378	0.90550
Zxd2:avg|sum	0.77624	2.92736	0.34160	0.90466
tsd:avg|sum	0.76554	2.99182	0.33426	0.90694
Zxd1:padding	0.77514	2.93590	0.34062	0.90822
tfidf:padding	0.77538	2.93682	0.34048	0.90664
Zxd2:padding	0.77514	2.93754	0.34042	0.90436
tsd:padding	0.76740	2.97964	0.33562	0.90834
Our model	0.77281	2.95292	0.33894	0.91132