计算机科学 ›› 2026, Vol. 53 ›› Issue (7): 298-307.doi: 10.11896/jsjkx.260200102
纪文迪, 王永全
JI Wendi, WANG Yongquan
摘要： 异常检测是医保基金监管中的关键技术。现有主流方法多依赖复杂的特征工程与业务专家经验来刻画可疑行为,检测规则的构建与维护成本高,且难以适应不断演化的欺诈模式。同时,医保异常标签稀缺、滞后且噪声较高,进一步制约了监督方法的可靠落地。为此,提出面向医保的无监督异常检测框架SeqRecon-AD,以医保账户的结算项目为基本单位,将账户历史按时间组织为项目序列,通过刻画序列在正常转移规律下的偏离程度来度量账户风险。模型采用自回归Transformer学习项目间的转移模式,以下一项目预测为训练目标,建模正常医保行为的序列规律。随后,计算位置级负对数似然损失,通过Top-k损失聚合得到账户级异常分数,从而突出少量关键异常片段的贡献,避免均值聚合对异常信号的稀释。在真实城市级数据集上进行了实验,结果表明,SeqRecon-AD优于经典无监督基线、代表性序列模型以及重构式自编码器方法,在不依赖训练异常标签的条件下提供了有效、可部署的无监督检测方案,相比最优无监督基线方法,其AUC提升了29.21%。
中图分类号:
| [1]LI Z,ZHU Y X,MATTHIJS V L.A survey on explainableanomaly detection[J].ACM Transactions on Knowledge Disco-very from Data,2023,18(1):1-54. [2]WANG Z Y,CHEN X F,WU Y W,et al A robust and interpretable ensemble machine learning model for predicting healthcare insurance fraud[J].Scientific Reports,2025,15:218. [3]LIANG C,LIU Z,LIU B,et al.Uncovering Insurance Fraud Conspiracy with Network Learning[C]//Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval.2019:1181-1184. [4]YANG S,ZHANG Z,ZHOU J,et al.Financial Risk Analysis for SMEs with Graph-based Supply Chain Mining[C]//Proceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence(IJCAI).2021:4661-4667. [5]EGRESSY B,VON NIEDERHÄUSERN L,BLANUŠA J,et al.Provably Powerful Graph Neural Networks for Directed Multigraphs[C]//Proceedings of the AAAI Conference on Artificial Intelligence.2024:11838-11846. [6]YOO Y,SHIN J,KYEONG S.Medicare Fraud Detection Using Graph Analysis:A Comparative Study of Machine Learning and Graph Neural Networks[J].IEEE Access,2023,11:88278-88294. [7]MA J,LI F,ZHANG R,et al.Fighting against Organized Fraud-sters Using Risk Diffusion-based Parallel Graph Neural Network[C]//Proceedings of the Thirty-Second International Joint Conference on Artificial Intelligence(IJCAI-23).2023:6138-6146. [8]WANG S,CAO J,YU P S.Deep Learning for Spatio-Temporal Data Mining:A Survey[J].IEEE Transactions on Knowledge and Data Engineering,2020,34(8):3681-3700. [9]HE Q,DING Q,ZHENG C,et al.A Data-Driven Intelligent Su-pervision System for Generating High-Risk Organized Fraud Clues in Medical Insurance Funds[J].Electronics,2025,14(16):3268. [10]SHI H,TAYEBI M A,PEI J,et al.Cost-Sensitive Learning for Medical Insurance Fraud Detection With Temporal Information[J].IEEE Transactions on Knowledge and Data Engineering,2023,35(10):10451-10463. [11]MAO Y,LI Y,XU B,et al.XGAN:A Medical Insurance Fraud Detector Based on GAN with XGBoost[J].Journal of Information Hiding and Multimedia Signal Processing,2024,15(1):36-52. [12]ZHANG R,CHENG D,YANG J,et al.Pre-trained Online Con-trastive Learning for Insurance Fraud Detection[C]//Procee-dings of the AAAI Conference on Artificial Intelligence.2024:22511-22519. [13]WU Y,ZHU Z,MA C,et al.Cost-Efficient Fraud Risk Optimization with Submodularity in Insurance Claim[C]//Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining.2024:3448-3459. [14]MA M,HAN L,ZHOU C.Research and Application of Transformer-based Anomaly Detec-tion Model:A Literature Review[J].arXiv:2402.08975,2024. [15]GUO H,YUAN S,WU X.LogBERT:Log Anomaly Detectionvia BERT[C]//2021 International Joint Conference on Neural Networks(IJCNN).IEEE,2021:1-8. [16]GUO L,LI R,JIANG B.A Data-Driven Long Time-Series Electrical Line Trip Fault Prediction Method Using an Improved Stacked-Informer Network[J].Sensors,2021,21(13):4466. [17]PINAYA W H L,TUDOSIU P D,GRAY R,et al.Unsupervised Brain Imaging 3D Anomaly Detection and Segmentation with Transformers[J].Medical Image Analysis,2022,79:102475. [18]CHEN N,TU H,DUAN X,et al.Semi-supervised Anomaly Detection of Multivariate Time Series Based on a Variational Autoencoder[J].Applied Intelligence,2023,53(5):6074-6098. [19]CHEN Z,CHEN D,ZHANG X,et al.Learning Graph Structures with Transformer for Multivariate Time-Series Anomaly Detection in IoT[J].IEEE Internet of Things Journal,2021,9(12):9179-9189. [20]ALI Z,HUANG Y,ULLAH I,et al.Deep Learning for Medication Recommendation:A Systematic Survey[J].Data Intelligence,2023,5(2):303-354. [21]MA L,ZHANG C,WANG Y,et al.ConCare:Personalized Clinical Feature Embedding via Capturing the Healthcare Context[C]//Proceedings of the AAAI Conference on Artificial Intelligence.2020:833-840. [22]WU R,QIU Z,JIANG J,et al.Conditional Generation Net for Medication Recommenda-tion[C]//Proceedings of the ACM Web Conference 2022.2022:935-945. [23]WU J,MEI X,MAO R,et al.TakeCare:A Temporal-Hierarchical Framework with Knowledge Fusion for Personalized Clinical Predictive Modeling [J].Information Fusion,2026,126:103620. [24]LIU F T,TING K M,ZHOU Z H.Isolation Forest[C]//2008 Eighth IEEE International Conference on Data Mining.IEEE,2008:413-422. [25]LI K L,HUANG H K,TIAN S F,et al.Improving One-Class SVM for Anomaly Detection [C]//Proceedings of the 2003 International Conference on Machine Learning and Cybernetics.IEEE,2003:3077-3081. [26]REN H,YE Z,LI Z.Anomaly Detection Based on a Dynamic Markov Model[J].Information Sciences,2017,411:52-65. [27]HIDASI B,KARATZOGLOU A,BALTRUNAS L,et al.Session-based Recommendations with Recurrent Neural Networks[C]//Proceedings of the International Conference on Learning Representations(ICLR).2016. [28]KANG W C,MCAULEY J.Self-Attentive Sequential Recom-mendation[C]//2018 IEEE International Conference on Data Mining(ICDM).IEEE,2018:197-206. [29]ZHAN K,WANG C,ZHENG X,et al.Seq2Seq-based GRU Autoencoder for Anomaly Detection and Failure Identification in Coal Mining Hydraulic Support Systems[J].Scientific Reports,2025,15(1):542. [30]KAMAL H,MASHALY M.AE-DTNN:Autoencoder-Dense-Transformer Neural Network Model for Efficient Anomaly-based Intrusion Detection Systems[J].Machine Learning and Knowledge Extraction,2025,7(3):78. |
|
||
首页