虛設公司預警系統之研究

A Study on the Early Warning System 0f Fictitious Company

林貴梅(Kuei-Mei Lin)

指導教授 ： 蔡碩倉

亞洲大學/管理學院/經營管理學系碩士班/碩士(2004年)

摘要

公司法之立法精神採準則主義，公司登記採書面形式審查，祇要申請人所送申請書件，符合公司法及有關法令規定，主管機關即應准其登記，使不肖之徒得以利用人頭或偽造身分證影本等申請虛設公司，向金融機構超貸，或利用各種名義吸金詐騙、洗錢等違法事件。本研究的目的除瞭解現行公司法相關規定，並探討虛設公司產生的原因與其衍生的問題，期能針對虛設公司申請設立、變更登記檢附的書件及其有無營業等進行事後追蹤探討，提出有效偵查虛設公司變數，以建立預警系統補足書面審查之不足，期能減少虛設公司之氾濫及對金融機構防止詐貸有所助益。
本研究所稱之「虛設公司」係指未依公司法規定登記設立者或其設立目的為存心不良，刻意虛設之公司組織，欲行詐欺詐騙者統稱之。對象為截至2004年底設籍於台灣省轄區實收資本額未達新台幣1億元之申請或已核准登記為有限公司及股份有限公司為資料來源，以公司法令、公司之登記及認許辦法、公司申請登記資本額查核辦法三個資料來源警訊為構面，共五項變數，應用倒傳遞類神經網路學習模式建構預警系統，並藉由敏感度分析比較各變數對於虛設公司影響程度。
經實證結果發現四種模式以網路Ⅳ（Norm-Cum-Delta-Rule為學習法則，Sigmoid為轉換函數），測試樣本正確分類率82.16﹪，誤差均方根RMSE=0.2174為最佳模式。另經由敏感度分析以「身分證被冒用偽造」最容易導致虛設公司的產生，其餘為「會計師資金查核簽證虛偽不實」、「資本股款實際繳納不實」、「無營業跡象」及「公司登記地址不實遭致退件」。

關鍵字

虛設公司 ； 倒傳遞類神經網路 ； 預警模型 ； 公司登記 ； 申請書 ； 身分證 ； 公司法 ； 敏感度 ； 申請人

並列摘要

The company law built is based on criterion, and the registration is checked by paper formula. As long as the apply paper accords to company law and to the related government provision, the responsible machinery organ should accept the registration. And, it may result in criminal to use the unaware individual or create fake ID to apply fictitious company. Then, with the fictitious company, they ask banks for the loan, or raise stock to cheat money. The objective for the research paper is to understand nowadays related company laws, and discuss the reasons fictitious company establish and find out the problems created by fictitious company. With the company application, register paper alteration, whether the company commenced its business...... and so on, I hope to offer effective variables to reconnoiter fictitious company. Further, to build early warning system to suffice the poverty of paper check, I hope to decrease fictitious company flow and help the financial organ to prevent cheat loan.
The so-called fictitious company here indicates a company established, and does not accord to company law and registration. Or its purpose is mean, and try to play deceit. The population tries to collect companies up to year of 2004 which instituted under Taiwan province, and their capital do not reach 100 millions but have been agreed and registered a limited company & company limited by shares for research data, it takes company law、registration & recognition of company and regulation governing the foregoing process capital of company applying registration for caution sign structure, and select five variables, apply into Back-propagation Network (BPN) Learning early warning system, and compare the influence of all variables on fictitious company by sensitivity analysis.
The experimental result has been found the best on four kinds of network structure is the network IV(Norm-Cum-Delta-Rule, Sigmoid): correct predict rate of the samples is up to 82.16%, and RMSE(Root Mean Square Error) is only 0.2174. By using sensitivity analysis, the most sensitive for fictitious company is 「the ID is fake or forged」. The others are 「an auditing certificate from an independent certified public accountant is not true」. 「the share price(or the capital stock) receivable by a company have not been actually paid up by its shareholders, but are declared as having paid up in its incorporation application」、「the company fails to or has discontinued commence its business operation at its own discretion for a period over six months」 and 「the document is sent back for the company registered fake address」.

並列關鍵字

fictitious company ； Back-propagation Network ； Early Warning System ； company register ； system ； taiwan ； data

參考文獻

翁景惠、高一書（2003）。測謊在我國法院使用之實證研究。台北市：刑事局鑑識中心。

陳志龍（2000）。法治國檢察官之偵查與檢查制度。台大法學論叢。27（3）。83。

臺灣省政府建設廳編印（1999）。臺灣省政府建設廳志。南投市。

鄭銌鋪（2004）。壽險業健康醫療險理賠預警系統之研究，臺中健康暨管理學院經營管理學系。台中市：未出版碩士論文。

Altman, E. I.（1968）. Financial Ratios, discriminant analysis,and the prediction of corporate bankruptcy. Journal of Finance.l（23）,. 589-609.