Table 3.

Clustering evaluation scores for different hyperparameter configurations on the MIMIC-IV dataset. (SS: Silhouette Score; CHS: Calinski-Harabasz Score; DBS: Davies-Bouldin Score). The configuration M=1, d=8, h=2, and K=3 achieved the best performance across all metrics, suggesting it as the optimal setup for the MIMIC-IV dataset.

Hyperparameters	K=3			K=4			K=5
	SS	CHS	DBS	SS	CHS	DBS	SS	CHS	DBS
M=1, d=8, h=2	0.25	884.95	1.47	0.24	680.51	1.67	0.13	458.59	1.89
M=1, d=8, h=4	0.23	631.74	1.78	0.13	707.22	1.87	0.16	671.80	1.72
M=1, d=16, h=2	0.17	457.76	2.02	0.13	547.68	2.43	0.14	463.17	2.26
M=1, d=16, h=4	0.13	455.67	2.41	0.10	460.91	2.41	0.10	478.54	2.20
M=1, d=32, h=4	0.08	280.06	2.81	0.11	269.57	2.99	0.09	257.99	2.85
M=1, d=32, h=8	0.05	228.10	3.36	0.06	197.85	3.12	0.05	180.37	3.15
M=2, d=8, h=2	0.14	810.25	2.06	0.16	609.06	1.89	0.16	669.63	1.77
M=2, d=8, h=4	0.20	680.40	1.87	0.12	691.99	1.85	0.22	799.93	1.55
M=2, d=16, h=2	0.09	342.08	2.65	0.09	332.93	2.40	0.13	449.23	2.20
M=2, d=16, h=4	0.13	463.70	2.19	0.09	296.01	2.44	0.09	274.68	2.41
M=2, d=32, h=2	0.11	358.12	2.69	0.07	246.35	2.92	0.06	162.41	3.38
M=2, d=32, h=4	0.10	311.41	2.82	0.07	237.12	2.94	0.06	196.55	3.34
M=2, d=64, h=4	0.13	438.68	2.47	0.12	338.72	2.91	0.12	329.10	2.51
M=2, d=128, h=4	0.14	416.29	2.85	0.14	368.75	2.32	0.14	346.96	2.34