Using Back-Propagation Neural Network and Support Vector Regression in Forecasting of Nikkei225 Stock Index and trading strategy

• Main Authors: Hsiang-Yu Liu, 劉翔瑜
• Other Authors: Tian-ShyugLee
• Format: Others
• Language: zh-TW
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/86544622591427509846

碩士 === 輔仁大學 === 金融研究所 === 94 === This study investigates the information content and spill over effect of Nikkei 225 futures prices during the non-cash-trading (NCT) period. The same day’s leading futures and previous day’s cash and futures market closing indices are firstly used to predict the opening cash price in the cash market by the back propagation neural network (BPN) and support vector regression (SVR) models. Sensitivity analysis is employed to address and solve the issue of finding the appropriate setup of the networks topology for both BPN and SVR. To demonstrate the effectiveness of our proposed method, the five-minute and one-minute intraday data of spot and futures index from September, 1998 to October, 2004 was evaluated using BPN and SVR. Analytic results demonstrate that the NCT futures prices do provide useful information in predicting the opening cash price index, the one-minute intraday data provide more information than the five-minute intraday data, and SVR has better prediction capability than BPN. Finally a proposed trading strategy using the observed results can provide significantly better investment return than the commonly discussed buy and hold strategy.