The effects of class rarity on the evaluation of supervised healthcare fraud detection models

Journal of Big Data

Table 10 Tukey’s HSD: learner

Part B				Part D				DMEPOS				Combined
Pos count	Ratio	AUC	Group	Pos count	Ratio	AUC	Group	Pos count	Ratio	AUC	Group	Pos count	Ratio	AUC	Group
(a) Train_Test
All	LR	0.82521	a	All	LR	0.79411	a	All	GBT	0.78241	a	All	LR	0.84996	a
1000	LR	0.82227	a	400	LR	0.77763	a	All	RF	0.78182	a	200	LR	0.80979	a
400	LR	0.81501	a	200	LR	0.75399	a	400	GBT	0.76752	a	100	LR	0.75512	a
200	LR	0.79216	a	100	LR	0.71832	a	400	LR	0.76307	a	All	GBT	0.83355	b
All	GBT	0.80481	b	All	GBT	0.76307	b	400	RF	0.76160	a	200	GBT	0.78286	b
1000	GBT	0.80152	b	400	GBT	0.73238	b	200	LR	0.73798	a	100	GBT	0.73082	b
400	GBT	0.78540	b	200	GBT	0.69234	b	100	LR	0.71709	a	100	RF	0.71527	b
200	GBT	0.76252	b	200	RF	0.68840	b	200	RF	0.73094	ab	All	RF	0.81608	c
All	RF	0.77776	c	100	RF	0.66739	b	All	LR	0.77132	b	200	RF	0.76825	c
1000	RF	0.77407	c	100	GBT	0.66406	b	200	GBT	0.72607	b	–	–	–	–
400	RF	0.76327	c	All	RF	0.74568	c	100	GBT	0.69519	b	–	–	–	–
200	RF	0.74408	c	400	RF	0.71116	c	100	RF	0.69174	b	–	–	–	–
(b) Train_CV
All	RF	0.81760	a	All	LR	0.78270	a	All	LR	0.73666	a	All	LR	0.81233	a
1000	LR	0.81068	a	400	LR	0.76768	a	400	LR	0.72964	a	All	RF	0.80893	a
400	LR	0.79119	a	200	LR	0.73768	a	200	LR	0.68823	a	200	LR	0.77946	a
200	LR	0.76530	a	100	LR	0.70019	a	100	LR	0.66918	a	100	RF	0.71356	a
All	LR	0.81424	b	All	GBT	0.75729	b	All	GBT	0.72763	b	100	LR	0.70868	a
1000	RF	0.80668	b	400	GBT	0.73009	b	All	RF	0.72404	b	100	GBT	0.70509	a
1000	GBT	0.80495	b	200	GBT	0.68435	b	400	GBT	0.71176	b	All	GBT	0.79702	b
400	GBT	0.78070	b	200	RF	0.68094	b	400	RF	0.70983	b	200	RF	0.76253	b
200	RF	0.74957	b	100	RF	0.65973	b	200	RF	0.66659	b	200	GBT	0.75880	b
200	GBT	0.74753	b	100	GBT	0.65659	b	200	GBT	0.66642	b	–	–	–	–
All	GBT	0.80981	c	All	RF	0.74213	c	100	GBT	0.64334	b	–	–	–	-
400	RF	0.77444	c	400	RF	0.71214	c	100	RF	0.64295	b	–	–	–	–