Medicare fraud detection using neural networks

Journal of Big Data

Table 18 Average focal loss results (30 runs)

Method	\(\alpha\)	\(\gamma\)	Hidden layers	Decision threshold	ROC AUC	TPR	TNR	G-Mean
FL-1-2	0.25	2	2	0.0315	0.8015	0.8741	0.5202	0.6722
FL-1-4			4	0.0315	0.8019	0.8167	0.6264	0.7115
FL-2-2	0.25	3	2	0.0730	0.8073	0.7616	0.7019	0.7295
FL-2-4			4	0.0730	0.8020	0.7912	0.6595	0.7184
FL-3-2	0.25	4	2	0.1195	0.8071	0.7342	0.7309	0.7310
FL-3-4			4	0.1190	0.8025	0.7769	0.6781	0.7230
FL-4-2	0.25	5	2	0.1615	0.8072	0.7574	0.7018	0.7267
FL-4-4			4	0.1615	0.8030	0.7646	0.6952	0.7267