整合數據驅動方法以進行稽核績效最佳化之稽核指派

An Integrated Data-Driven Methodology for Auditor Performance Appraisals and Auditor Assignment Optimization

近年來，隨著金融業相關事業群擴張及產業數位化，金控公司常會因稽核人力調度的困難，而造成內部監督管理疏失，導致裁罰案之發生。本研究探討績效評估方法論與產業稽核績效評估模式，並進一步整合隨機森林和差分進化技術，提出二階段數據分析模型；透過分析金控稽核數據來最佳化稽核人員績效值，並解決稽核績效評估與任務規劃問題。驗證結果顯示：(1)二階段數據分析模型準確地評估稽核人員在不同工作下稽核表現，指派合適稽核任務，強化運營決策；(2)建構專業課程、稽核資歷、稽核任務等構面的多目標決策模型，分類最佳方案及有應用價值的派工法則，對稽核人力管理極有助益；(3)稽核數據專案的實地運用中，啟發式方法會較傳統規劃法更具可執行性。

With the expansion of related business groups and impact of digitization on financial industry in Taiwan, in recent years, financial holding companies often fail to provide adequate internal supervision and management due to problems of audit-staff scheduling, which frequently leads to adjudication and penalization. The study aims to solve the problem and find better performance evaluation method of auditors beginning by reviewing past performance appraisal models in prior literature. We then integrate the random forest and differential evolution technology and propose a two-stage data analysis method. Applying this method, we analyze audit data collected from the financial information system; optimize the internal audit performance; and resolve issues related to audio performance evaluation and task planning. The verification results show this two-stage model can accurately evaluate the audit performance of high-level audits under different tasks; smoothly assign appropriate audit tasks; and strengthen operational decisions. Secondly, we prove the construction of a multi-objective mathematical model based on provision of professional courses, audit qualification system, task auditing, etc. can optimally classify the most applicable action plan and dispatch rules, which are extremely helpful to the management of audit-staff scheduling. Lastly, with the practical application of this audit data project, we demonstrate the heuristic method suggested would be more feasible in execution than the conventional planning method.