Leveraging feature selection to detect potential tax fraudsters
作者：
Highlights：
• Novel feature selection ranking method in the context of tax fraud detection.
• Our method outperforms its competitors, achieving F-measure scores up to 76.88%.
• Set of real-world datasets are used for validation.
• The intrinsic interrelation between features is captured in a Graph.
• New Centrality Measure and novel axiomatic description of dependency structures.
摘要
•Novel feature selection ranking method in the context of tax fraud detection.•Our method outperforms its competitors, achieving F-measure scores up to 76.88%.•Set of real-world datasets are used for validation.•The intrinsic interrelation between features is captured in a Graph.•New Centrality Measure and novel axiomatic description of dependency structures.
论文关键词：Feature selection,Tax fraud detection,Association rules,Graph centrality measure
论文评审过程：Received 3 September 2018, Revised 10 November 2019, Accepted 8 December 2019, Available online 9 December 2019, Version of Record 27 December 2019.
论文官网地址：https://doi.org/10.1016/j.eswa.2019.113128
