Explainable Fraud Detection for Few Labeled Time Series Data

Table 3

The experimental results on the PaySim simulation dataset with missing labels.

		(%)	Recall (%)	Precision (%)	AUC
AP	0.4233	60.49 ± 3.74	32.28	61.46	0.6083
SVM	0.5911	51.57 ± 2.42	25.32	54.13	0.5275
RF	0.6837	58.96 ± 2.47	31.22	61.46	0.5982
HOBA	0.3543	53.46 ± 1.74	54.12	27.18	0.5346
Ours	0.7789	84.58 ± 2.27	61.92	86.69	0.8511