摘 要
在現今全球金融環境變化快速的時代裡，企業愈來愈仰賴大量的電腦資訊以維持競爭力，近來類神經網路的應用漸漸被使用在提高財務決策之品質與效益，以作為財務決策支援系統。類神經網路在財務方面之應用相當廣泛，由信用評等到許多財務上之預測，例如分析客戶資料庫以找尋潛在客戶群以及分析財務交易資料庫以找尋洗錢情形等等。因此，企業在龐大的資料庫中，運用由類神經網路所作之決策支援系統，將這些原先無法得知的資訊，用來增加獲利、加強客戶服務，最後將增加企業競爭力。
類神經網路有一項重要的特徵，就是在資料不完全的情況下，類神經網路有能力作出合理的解釋。財務資料常常含有許多雜訊或是財務資料不完整的情況出現，因此類神經網路的特性在財務領域上非常重要，這與多變數迴歸分析法對所有的變數須使用嚴謹的檢定後，才能使迴歸分析的結果具有解釋能力大不相同。
IPO是一個公開發行公司首次公開賣出股票的時機，公司得以由資本市場取得營運資金。而投資者如何評估一個公司的IPO價格是否允當，則需依賴許多的變數，加上它們彼此獨立而且關係不明確，這情況使得投資者或投資銀行不易決定IPO價格，此外，在決定IPO價格時只能依據不完全的資訊。本研究依據類神經網路的特性，從法人機構或投資銀行的角度，使用類神經網路對IPO蜜月期收盤價作預測，蜜月期收盤價是指經過數天漲停或跌停被打開之前一天的收盤價。類神經網路的預測績效將與多變數迴歸模型相比較，結果顯示，類神經網路預測值之相關係數及平均差異比率均優於多變數迴歸模型預測值之相關係數及平均差異比率，故類神經網路對IPO後之蜜月期結束之收盤價之預測能力較佳。

Abstract
Financial Service Companies are getting more dependent on using computers to maintain competitiveness at today’s rapidly changing financial environment. Artificial neural networks are getting more popular as a tool to aid in the quality of financial applications. Artificial neural network has found its way into many sectors of financial application, ranging from credit card authorization to various financial forecasting. Examples like analyzing database to find potential customers, and sip through database to isolate money-laundry activities. With this tool as a decision support system, corporation can find knowledge never found before and to use it to gain better benefit, enhance customer service, and increase competitiveness.
One characteristics of artificial neural network is even with incomplete or imperfect data, it can still provide reasonable answer. And this is very important for financial forecasts since financial data are sometimes incomplete or containing erroneous data. And contrary to the strict data cleaning requirement for multiple regression technique to come up with a good answer,
IPO is the first chance a corporation can garner public capital from the capital market by selling its stock to the market. How to evaluate the appropriate price for the post IPO price needs not only information regarding the subject company, but also the economic environment, also the input variables are independent no relation with each other, causing individual investors or investment bank hard to determine the post IPO price. And add to fact that available information is sometimes incomplete , it is even harder to do proper valuation.
This report utilizes the characteristics of artificial neural network, from the perspective of an individual investor or investment bank, so with limited financial data available from the market, to do post IPO price forecast. While we define the post IPO price is the price of the previous day before the 7% up or down limitation lock is broken. The result of the artificial neural network is compared with the result from the multiple regression technique. And the conclusion is that the average deviation from artificial neural network is substantially lower then those from the multiple regression method, so artificial neural network has better performance in the case of post IPO price.

目 錄
第1章 緒論 1
1.1 研究動機 1
1.2 研究目的 3
1.3 研究假說 4
1.4 研究範圍與研究限制 5
1.5 研究架構與流程 7
第2章 文獻回顧 11
2.1 類神經網路 11
2.2 人工類神經網路(ARTIFICIAL NEURAL NETWORK) 12
2.3 訓練網路 20
2.4 倒傳遞(BACK PROPAGATION)類神經網路 21
2.5 類神經網路應用之相關文獻 24
第3章 研究方法 28
3.1 類神經網路設計流程 28
3.1.1 收集資料 30
3.1.2 資料分析 38
3.1.3 網路設計 40
3.2 網路訓練與測試 43
3.2.1 網路訓練 43
3.2.2 網路測試 44
第4章 實證研究 45
4.1 評價方法 45
4.2 多變數迴歸實證 46
4.2.1 資料分析 46
4.2.2 迴歸結果 49
4.3 類神經網路實證 50
4.3.1 訓練集與測試集資料選定 50
4.4 模型評量 59
第5章 結論 60
5.1 驗證假說之結果 60
5.2 結論與建議 63
參考文獻 64
附錄A 增減神經元測試 68
附錄B 調整學習參數測試 73
附錄C 隱藏層增加到2層之測試 76
附錄D 原始資料 78

一、中文部分
1.吳振坤，81年6月，類經神網路在學習臺灣股價行為上的應用，交通大學資訊管理研究所碩士論文。
2.周育蔚，85年6月，利用類經神網路建立台灣股價預測模型，台灣大學商研所碩士論文。
3.林志鴻，84年6月，類經神網路支援股市投資決策，台灣大學商研所碩士論文。
4.張文信，84年6月，以類經神網路預測股價指數漲跌，台灣大學財務金融所碩士論文。
5.陳玉清，84年6月，以類神經網路建立股票價格預測知識庫，東吳大學會計系碩士論文。
6.葉怡成，84年4月，類神經網路模式應用與實作，儒林書局。
7.盧炳勳，曹登發，81年3月，類神經網路理論與應用，全華書局。
8.詹錦宏、袁澤峻，88年10月，景氣循環與股價指數之預測─類神經網路之應用， 證交資料450期。
9.許蕙婷，84年6月，我國創業投資事業參與行為及價值貢獻之研究，中山大學財務管理系碩士論文。
10.方伯中，85年6月，我國創投對被投資公司IPO之影響及未來發展方向，台灣大學財務金融學研究所碩士論文。
11.劉松瑜，90年7月，創業投資公司投資行為對被投資公司績效影響之研究，國立政治大學企業管理學系博士論文。
二、英文部分
1. Baba, N.and M. Kozaki, 1992, “An Intelligent Forecasting System of Stock Price Using Neural Networks”, IEEE/INNS International Joint Conference on Neural Networks-Baltimore, PⅡI371-I377.
2. Donaldson,R.G. and M. Kamstra, 1996, “Forecast Combining with Neural Networks”, Journal of Forecasting, Vol.15, P49-61.
3. Fishman, M.B., D.S. Barr, and W.J. Loick, 1991, “Using Neural Nets in Market Analysis”, Technical Analysis of Stocks and Commodities, Vol.9, P18-21.
4. Gately ,E., 1996,Neural Networks for Financial Forecasting, John Wiley & Sons. Inc.,
5. Gencay,R., 1996, “Non-linear Prediction of Security Returns with Moving Average Rules”, Journal of Forecasting, Vol.15, P165-174.
6. Hawley, D.and D. Raina, 1990, “Artificial Neural System: A New Tool for Financial Decision-Making”, Financial Analysts Journal, November/December,P63-72.
7. Haykin,S., 1994,Neural Network A Comprehensive Foundation, Macmillan Publishing Company.
8. Hill, T.,M. O’Connor and W. Remus, 1996, “Neural Network Models for Time Series Forecasts”, Management Science, Vol.42, No7, P1082.
9. Kamijo, K.and T. Tanigawa, 1990, “Stock Price Pattern Recognition: A Recurrent Neural Network Approach”, IEEE International Joint Conference on Neural Networks PⅠ215-221.
10. Kimoto, T.and K. Asakawa, 1990, “Stock Market Prediction System with Modular Neural Networks”, IEEE International Joint Conference on Neural Networks PⅠ1-I6.
11. Kuan, C. andH. White, 1994, “Artificial Neural Networks: An Econometric Perspective” ,Econometric Reviews, Vol.13 (1), P1-91.
12. Medsker, L.,E. Turban and R.R. Trippi, 1993, “Neural Network Fundamentals for Financial Analysts”, collected in Neural Networks in Finance and Investing, Probus Publishing Company, P3-25.
13. NeuroDimension, Inc. 1996,Neuro Solutions Getting Started Manual Version 3.0,Neuro Dimension, Inc.
14. Remelhart,D.E., G.E. Hiton, and R.J. Williams, 1986, “Learning Internal Representations by Error Propagation”, collected in Parallel Distributed Processing: Explorations in the Microstructure of Cognition, Remelhart, D.E. and J.L.McClelland, eds., Vol.1, MIT Press.
15. Rosenblatt,F., 1962,Principle of Neuro dynamics, Spartan Books.
16. Skapura,D.M., 1995,Building Neural Networks, Addison-Wesley Publishing Company.
17. Surkan, A.J.and A.N. Skurikhin, 1996, “Experiments in Bond Rating with Probabilistic Neural Networks”, Progress in Neural Information Processing, Vol.2: ICONIP’96(Hong Kong), P705.
18. Werbos, P.J.,1974, Beyond regression: New tools for prediction and analysis in the behavioral sciences, Ph.D. Dissertation, Harvard University, Cambridge, MA.
19. White, H.,1988, “Economic Prediction Using Neural Networks: The Case of IBM Daily StockReturns”, IEEE International Joint Conference on Neural Networks, PⅡ451-458.
20. White, H., K.Hornik and M. Stinchcombe, 1992,Artificial Neural Networks, Blackwell Publishers.
21. Trajtenberg, Manuel, Jan2002, Government Support for Commercial R&D: Lessons from the Israeli Experience, NBER Innovation Policy & the Economy, Vol. 2 Issue 1, p79.
22. Abravanel, Roger , 2001, The promised economy, McKinsey Quarterly, Issue 4, p133.
23. Paul A. Gompers and Josh Lerner , 1999, The Venture Capital Cycle, The MIT Press.