Application of Outlier Detection Methods in Audit Analytics
研究了如何将无监督学习的异常值检测方法用于审计，提出一个框架帮助审计人员从交易数据中识别风险点，并在真实收入明细数据上验证了效果。
SYNOPSIS Audit transaction anomalies can be viewed as outliers. Unsupervised learning methods of outlier detection do not require outcome labels and enable auditors to discover possible problems based on observed transaction patterns. This study develops a framework for using outlier detection methods in audit selection and evaluates the proposed framework on real-world revenue subledger datasets. The results indicate that the proposed framework could facilitate the identification of relevant outlier detection algorithms and effectively select risky observations.