Medicare fraud detection using neural networks

Journal of Big Data

Table 13 Average RUS results (30 runs)

Method	\(n_{neg}{:}n_{pos}\)	Hidden layers	Decision threshold	ROC AUC	TPR	TNR	G-Mean
RUS-1-2	99:1	2	0.0110	0.8124	0.7807	0.6987	0.7383
RUS-1-4	99:1	4	0.0145	0.8040	0.7581	0.7002	0.7265
RUS-2-2	80:20	2	0.2680	0.8076	0.7521	0.7163	0.7338
RUS-2-4	80:20	4	0.3520	0.7920	0.7674	0.6853	0.7228
RUS-3-2	60:40	2	0.4200	0.8043	0.7783	0.6700	0.7212
RUS-3-4	60:40	4	0.5370	0.7907	0.7978	0.6288	0.7021
RUS-4-2	50:50	2	0.4970	0.8027	0.7864	0.6601	0.7195
RUS-4-4	50:50	4	0.6078	0.7913	0.7778	0.6422	0.6966
RUS-5-2	40:60	2	0.5730	0.7994	0.7802	0.6588	0.7154
RUS-5-4	40:60	4	0.7060	0.7802	0.7226	0.6462	0.6412