Applied Network Science

Table 6 Average area under the curve (<AUC>) values and Standard Deviations (<STD>) for each complication of diabetes based upon the number of significant nodes and edges that were included in the prediction

From: Predicting onset of complications from diabetes: a graph based approach

Nodes/Edges	STR	RET	KID	MYO	HFL	LIV	<AUC>	<STD>
7	0.767	0.793	0.762	0.734	0.766	0.733	0.759	0.021
10	0.785	0.788	0.760	0.748	0.771	0.724	0.763	0.022
12	0.780	0.778	0.755	0.743	0.776	0.743	0.763	0.016
15	0.762	0.792	0.732	0.751	0.764	0.769	0.762	0.018
17	0.743	0.772	0.757	0.757	0.767	0.724	0.753	0.016
20	0.797	0.761	0.742	0.725	0.762	0.740	0.755	0.023

We see that using the 12 most significant nodes and edges gives us the highest average AUC with the lowest standard deviation across the various complications. Abbreviations: Myocardial infarction (MYO), Heart failure (HFL), Kidney disease (KID), Liver disease (LIV), Retinopathy (RET), Stroke (STR)

Back to article page