運用類神經網路建構指數套利模型-以日經225指數為例

Abstract

指數套利行為的產生，乃是期貨實際價格與期貨理論價格間之乖離過大所 造成，但是，此乖離過大，經常為一偶發的事情。因此，能否快速而有效 率的達成套利策略所應有的部位，將決定整個套利活動是否真正為無風險 套利(Risk-free arbitrage)，否則，套利行為將極具風險。本研究即希 望利用類神經網路作為套利模型之行為分析工具，以研判基差之走勢，並 建構較佳的指數套利模型。本研究以日經225股價指數之市場為樣本，分 別以持有成本套利模型為對照組，及運用類神經網路作為預測工具的類神 經網路套利模型為實驗組，進行指數套利獲利分析。 本研究結果 發現：1. 基差擴大為一持續性的現象，顯示執行套利時，值得以預測基 差走勢之方式來進行。2. 以類神經網路套利模型執行指數套利時，將可 比持有成本套利模型多獲取兩倍以上的套利利潤。3. 類神經網路套利模 型在方向性波動較大的T_基差變動下，會忽略反轉點的趨勢；雖然會錯失 利潤較少的套利機會，但將會減低市場衝擊風險。4將日經225指數期貨與 現貨市場之價量技術指標作資料前處理後，使資料對問題本身更具意義。 The cause of index arbitrage occurrence is the over basis between future practice price and theoretical price, but it happens suddenly. Thus, arbitrageur make the position efficiency and quickly will decide the whole index arbitrage actions is risk-free or not. The Research uses the Neural Networks to be the analysis tool for index arbitrage model and uses the Nikkei 225 stock index for sample data. The thesis constructs cost-of-carry model and uses the Neural Networks to be forecasting tool for Neural Network arbitrage model. The study analyzes the return of these two index arbitrage models. The results of this study are as following:(1) The basis is wild continually, so we may forecast the trend of index basis.(2) If we use Neural Networks within index arbitrage model will get over double return than cost-of-carry arbitrage model.(3) If the T_basis has volatile trend, the Neural Networks arbitrage model will ignore the peak. Although arbitrageur would lost the chance to get profit, they may reduce the market impact risk.(4) It is more useful to preprocess the technique indexes of Nikkei 225.