Topological pattern discovery and feature extraction for fraudulent financial reporting

作者：

Highlights：

• We discover the spatial relationships of fraud and non-fraud financial statements.

• An expert-competitive feature extraction mechanism is presented to capture the salient characteristics of fraud behaviors.

• We design a parameter adjustment method to develop the classifiers in topological space.

• The proposed method can generate classification rules to help detect fraudulent samples.

摘要

•We discover the spatial relationships of fraud and non-fraud financial statements.•An expert-competitive feature extraction mechanism is presented to capture the salient characteristics of fraud behaviors.•We design a parameter adjustment method to develop the classifiers in topological space.•The proposed method can generate classification rules to help detect fraudulent samples.

论文关键词：Unsupervised learning,Growing Hierarchical Self-Organizing Map,Data mining,Fraudulent financial reporting

论文评审过程：Available online 24 January 2014.

论文官网地址：https://doi.org/10.1016/j.eswa.2014.01.012