Mining corporate annual reports for intelligent detection of financial statement fraud – A comparative study of machine learning methods
作者：
Highlights：
• We combine features derived from financial information and managerial comments.
• We employ feature selection and classification using a wide range of machine learning methods.
• Analysts’ forecasts of revenues and earnings are necessary to detect fraudulent firms.
• Misclassification cost ratio of 1:2 is based on the loss attributable to financial statement fraud and audit fees.
• Interpretable Naïve Bayes-based models outperform remaining methods in terms of misclassification costs.
摘要
•We combine features derived from financial information and managerial comments.•We employ feature selection and classification using a wide range of machine learning methods.•Analysts’ forecasts of revenues and earnings are necessary to detect fraudulent firms.•Misclassification cost ratio of 1:2 is based on the loss attributable to financial statement fraud and audit fees.•Interpretable Naïve Bayes-based models outperform remaining methods in terms of misclassification costs.
论文关键词：Annual reports,Financial statement fraud,Text mining,Feature selection,Machine learning
论文评审过程：Received 5 December 2016, Revised 27 February 2017, Accepted 2 May 2017, Available online 6 May 2017, Version of Record 25 May 2017.
论文官网地址：https://doi.org/10.1016/j.knosys.2017.05.001
