類神經模糊專家系統在訴訟預警模型之應用：以公司治理觀點

Abstract

本研究著眼於以公司治理觀點，來建構類神經模糊專家系統之訴訟預警模式，藉以輔助審計人員判斷受查客戶之風險程度，並進一步瞭解風險之由來。以傳統的羅吉斯迴歸作為類神經模糊專家系統解釋與預測績效的比較基準。實證結果發現，類神經模糊相較於傳統預測工具能提供較高的預測準確度、檢定力與資訊價值；而類神經模糊所提供的知識庫規則對變數間關係有更為細膩之描述，其有利於未來假設的更進一步驗證。

The main purpose of this study is to construct a litigation-presence warning model based on the corporate governance factors by applying neuro-fuzzy expert system. This fuzzy expert system can not only help the auditors evaluate the audit risks but also provide the reasons behind. Compared with the traditional used logit model, the proposed neuron-fuzzy model can provide more accurate prediction, testing power and information value. Besides, the relationship among the variables described in the knowledge base is much more elaborate than that contained in the logit model, which can be further pursued to do the hypothesis testing and theory building.