深度學習技術於財報舞弊議題適用性之探討

CCUR > College of Business > Department of Accounting & Graduate Institute of Accounting > Thesis > Item 987654321/52987

Please use this identifier to cite or link to this item: https://irlib.pccu.edu.tw/handle/987654321/52987

Title:	深度學習技術於財報舞弊議題適用性之探討 Exploring the Applicability of Deep Learning Techniques in Financial Fraud Issues
Authors:	鄒永恩
Contributors:	會計學系
Keywords:	深度學習財報舞弊預測決策樹CART 類神經網路循環神經網路
Date:	2023
Issue Date:	2023-10-04 14:06:42 (UTC+8)
Abstract:	近年來財報舞弊事件發生頻繁，許多公司皆因財報舞弊事件導致下市，導致員工及投資人莫大的損害，也造成經濟市場的恐慌。財報舞弊的評估具複雜性，因此，如何能提早預測財報舞弊的發生，對於建立一個有效且準確的財報舞弊預測模型就相當重要。研究樣本取自於臺灣經濟新報資料庫(TEJ)，研究樣本為2000年至2022年台灣有發生財務危機之上市櫃公司，並以 1:3 比例進行樣本配對，變數方面包含 19 個財務變數及 12 個非財務變數。本研究以決策樹 CART 和羅吉斯進行篩選，再利用卷積神經網路(CNN)及循環神經網路(RNN)建立有效之財務危機預測模型。實證結果顯示，以決策樹 CART 搭配卷積神經網路在預測準確率最高，平均準確率達 96.41% In recent years, there has been a frequent occurrence of financial statement fraud, leading many companies to be delisted as a result. This has caused significant damages to employees and investors, as well as causing panic in the economic market. Assessing financial statement fraud is a complex task, making it crucial to find ways to predict its occurrence in advance. Therefore, establishing an effective and accurate financial statement fraud prediction model is highly important. The research sample for this study was sourced from the Taiwan Economic Journal (TEJ) database. It consisted of listed and OTC companies in Taiwan that experienced financial crises from 2000 to 2022. The sample was paired at a ratio of 1:3, and the variables included 19 financial variables and 12 non-financial variables. The research employed CART decision tree and logistic regression for screening, followed by the utilization of Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN) to construct an effective financial crisis prediction model. The empirical results showed that the combination of CART decision tree and CNN achieved the highest prediction accuracy, with an average accuracy rate of 96.41%.
Appears in Collections:	[Department of Accounting & Graduate Institute of Accounting] Thesis

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	137	View/Open

View Licence

Loading...