Data-driven fraud detection in international shipping

作者：

Highlights：

• A Bayesian network is developed to detect miscoding and smuggling from shipment data.

• Probabilistic discriminative models are derived from the Bayesian network.

• Data-driven fraud detection provides higher quality fraud alarms than random audits.

• Discriminative models tend to generate higher quality alarms than generative models.

摘要

•A Bayesian network is developed to detect miscoding and smuggling from shipment data.•Probabilistic discriminative models are derived from the Bayesian network.•Data-driven fraud detection provides higher quality fraud alarms than random audits.•Discriminative models tend to generate higher quality alarms than generative models.

论文关键词：International shipping,Fraud detection,Probablistic classification,Bayesian networks,Logistic regression,Neural networks

论文评审过程：Received 29 March 2017, Revised 30 November 2017, Accepted 7 January 2018, Available online 31 January 2018, Version of Record 3 February 2018.

论文官网地址：https://doi.org/10.1016/j.eswa.2018.01.007