An Evolutionary Computation Approach for Optimizing Multilevel Data to Predict Patient Outcomes

Table 5

Comparison of results generated from flattened and hierarchical approaches across five patient populations.

(a)

	Flattened
	ACAD	COMM	BRAZIL	UAE	NHAMCS
Overall AUC	0.8431	0.8361	0.8261	0.8820	0.8429
Training time (hr)	42.47	78.67	19.89	15.00	29.06
Selected complaints (%)	48.3	52.8	53.4	59.0	49.9

(b)

	Hierarchical
	ACAD	COMM	BRAZIL	UAE	NHAMCS
Overall AUC	0.8433	0.8364	0.8260	0.8819	0.8436
Training time (hr)	4.93	8.91	3.46	3.27	3.09
Selected complaints (%)	49.3	64.6	55.6	55.6	46.4

(c)

	Comparison
Difference in overall AUC ( value)	0.6144	0.2210	0.7022	0.3622	0.2579
Jointly selected complaints (%)	28.1	33.1	32.4	37.5	27.5
Jointly excluded complaints (%)	30.6	27.6	23.5	22.9	31.2