Please wait a minute...
Shopping Cart Email List Chinese
Advanced Search Fig/Tab
Frontiers of Computer Science

ISSN 2095-2228

ISSN 2095-2236(Online)

CN 10-1014/TP

Postal Subscription Code 80-970

2018 Impact Factor: 1.129

Featured Articles  |  Most Downloaded  |  Most Reading
Online Submission Subscribe to Our RSS Content Feed
Front. Comput. Sci.    2023, Vol. 17 Issue (2) : 172341    https://doi.org/10.1007/s11704-022-1734-0
LETTER
A graph-based contrastive learning framework for medicare insurance fraud detection
Song XIAO, Ting BAI(), Xiangchong CUI, Bin WU, Xinkai MENG, Bai WANG
School of Computer Science, Beijing University of Posts and Telecommunications, Beijing 100876, China
 Download: PDF(819 KB)   HTML
 Export: BibTeX | EndNote | Reference Manager | ProCite | RefWorks
Corresponding Author(s): Ting BAI   
Just Accepted Date: 15 November 2022   Issue Date: 17 January 2023
 Cite this article:   
Song XIAO,Ting BAI,Xiangchong CUI, et al. A graph-based contrastive learning framework for medicare insurance fraud detection[J]. Front. Comput. Sci., 2023, 17(2): 172341.
 URL:  
https://academic.hep.com.cn/fcs/EN/10.1007/s11704-022-1734-0
https://academic.hep.com.cn/fcs/EN/Y2023/V17/I2/172341
Fig.1  Contrastive learning can enhance fraud detection performance by distinguishing dissimilar diagnosis and gathering similar diagnosis (e.g., CAD and Hypertension, both of which are a kind of cardiovascular disease)
Fig.2  The overall framework of our proposed model. And the dashed box indicates treatment procedures with similar diagnosis
Fig.3  Problem formulation of medicare insurance fraud detection
Part I Part II Part III Average
F1-score AUC F1-score AUC F1-score AUC F1-score AUC
Linear regression 0.5412 0.7938 0.5309 0.7867 0.5250 0.7682 0.5324 0.7829
Decision tree 0.5723 0.8402 0.5541 0.8138 0.5531 0.8177 0.5589 0.8239
LSTM 0.7228 0.9788 0.7486 0.9833 0.7403 0.9850 0.7372 0.9824
ON-LSTM [3] 0.7521 0.9791 0.7661 0.9822 0.7582 0.9861 0.7588 0.9824
TLSTM [4] 0.7781 0.9743 0.7716 0.9809 0.7531 0.9778 0.7676 0.9777
VS-GRU [5] 0.7832 0.9748 0.7698 0.9812 0.7543 0.9745 0.7691 0.9768
HAInt-LSTM [2] 0.7854 0.9821 0.7642 0.9851 0.7783 0.9825 0.7760 0.9832
GCLF(ours) 0.8686 0.9913 0.8889 0.9977 0.8866 0.9975 0.8813 0.9955
Tab.1  Performances of different methods on the dataset
Dataset Type #Positive #Negtive #Positive rate
Part I Training 4,726 118,302 3.84%
test 501 14,312 3.38%
Part II Training 4,732 118,032 3.85%
test 514 14,541 3.53%
Part III Training 4,862 119,812 3.90%
test 541 13,989 3.72%
Tab.2  The statistical information of dataset
Part I Part II Part III
GCLF 0.8686 0.8889 0.8866
-Medicine graph 0.7922 0.7867 0.7987
-Contrastive loss 0.8487 0.8657 0.8583
Tab.3  F1-score of ablation test on the proposed GCLF
1 J, Li Q L, Lan E Y, Zhu Y, Xu D Zhu . A study of health insurance fraud in China and recommendations for fraud detection and prevention. Journal of Organizational and End User Computing, 2022, 34( 4): 1–19
2 Guo J, Liu G N, Zuo Y, Wu J J. Learning sequential behavior representations for fraud detection. In: Proceedings of 2018 IEEE International Conference on Data Mining. 2018, 127−136
3 Shen Y K, Tan S, Sordoni A, Courville A. Ordered neurons: integrating tree structures into recurrent neural networks. In: Proceedings of the7th International Conference on Learning Representations. 2019
4 Cao L H, Qin F L, Yan Z M. TLSTM-based medical insurance fraud detection. Computer Engineering and Applications, 2020, 56(21): 237-241
5 Q T, Li Y Xu . VS-GRU: a variable sensitive gated recurrent neural network for multivariate time series with massive missing values. Applied Sciences, 2019, 9( 15): 3041
[1] FCS-21734-OF-SX_suppl_1 Download
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed