標題: | 運用類神經網路預測選擇權評價模式中股票價格波動率之實證研究 An Empirical Study on Forecasting the Stock Volatility in the Option Pricing Model by Using the Neural Networks |
作者: | 鍾澄吉 Chung, Chenz-Chi 陳安斌 An-Pin Chen 資訊管理研究所 |
關鍵字: | 實證研究;Black-Scholes評價模式;價格波動率;預測;類神經網路;基因演算法;Empirical Study;Black-Scholes Pricing Model;Forecast;Volatility;Genetic Algorithm;Neural Network |
公開日期: | 1997 |
摘要: | 傳統上, Black-Scholes 評價模式是一種便捷的選擇權評價方法。 然而,在 Black-Scholes 評價模式中許多假設並不符合實際,因此所評 估出來的價格與實際價格亦有乖離現象。探究評價模型的所有輸入變數中 ,價格波動率是最難以瞭解且影響評價結果甚巨的因子。但標的物價格的 未來波動率是一項未知數。因此若要運用評價模型,就必須透過某種方法 預測契約期間內標的物價格的波動率。 然而近年來在人工智慧領域的 研究上,發現類神經網路具有學習機制與高速計算之特性,更由於類神經 網路之非線性處理與容錯的強大能力,使其預測的能力相當優異。因此, 本研究將針對類神經網路就一隨機過程之資料能有效掌握其波動行為的特 性,用以預測選擇權評價模式中標的物價格的未來波動率,並就其結果進 行分析與比較。 本研究針對30家台灣股票上市公司以模擬選擇權的交 易方式進行實證研究,嘗試利用基因演算法自動演化之類神經網路來架構 一價格波動率的預測模式,並與傳統上利用歷史波動率的模式相比較,且 相互評估對市場真實價格波動率的誤差程度與解釋能力。研究期間涵 蓋1995年至1997年,共計3年。 研究結果顯示,類神經網路預測模式 不論在解釋能力或實際避險策略操作上都優於傳統上所利用的方法。表示 基因演算法自動演化之類神經網路能提供一個較傳統模式更優良的方式, 以準確預估價格波動率,期能輔助投資者與發行方做出正確的投資決策與 避險策略。 Traditionally, the Black-Scholes model is a useable evaluation method for optionpricing. However, there exist some impractical assumptions in the Black-Scholes Model. Therefore, the evaluated prices would be reconsidered that may generated from the practiced data. Usually, in all input variables considered from B-S model, the volatility is the most difficulty part to understand. Thus, if a practical model would be applied, the volatility should be precisely estimated as the first step to be done.However recent studies of the field in the artificial intelligence reflect that neural networks have the ability of learning and performing high-speed calculations. Also with it's parallel processing and tolerance of faults, it's prediction ability has become gradually an accepted level. So this research will attempt to forecast the future volatility of prices by neural networks.This empirical research is to simulate options transactions that randomly selected from 30 companies in Taiwan. In this research, the neural network with genetic algorithm should be applied to forecast the volatility of each stock and the result would be compared with the output from the traditional B-S method. The research periods is starting from 1995 to 1997 three years.The research shows that the output from the forecast module by the neural network is superior to that from the traditional B-S method no matter in interpreter capability or in practical hedge strategy operation. This means that the new method provided in this thesis can assist the investor to make more precisely investment decision and hedge strategy. |
URI: | http://140.113.39.130/cdrfb3/record/nctu/#NT860396017 http://hdl.handle.net/11536/62971 |
Appears in Collections: | Thesis |