Applying Neural Networks and Genetic Algorithms in Taiwan Stock Index Futures Prediction of Opening Gaps

• Main Authors: Hsu, Shih Hang, 徐士航
• Other Authors: Lin,Wei Yuan
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/27428638368378991749

碩士 === 東吳大學 === 經濟學系 === 101 === The purpose of this study is to use artificial intelligence to Taiwan Futures Exchange's index (TAIFEX), it tries to combine Genetic Algorithm(GA) with Back-Propagation Neural Network(BPN) to predict the next day opening gaps direction of TAIFEX. In this study, we apply traditional regression, Logistic Regression,Genetic Algorithmand Back Propagation Neural Network models to analyze our evaluation to the next day opening gaps upside or downside in the market. Our algorithms started with original 41 variables, and 10 variables selected from Genetic Algorithm, then apply Matlab & Pythia programs to build six kinds of models with two sets of variables. The simulation results of these models with four methods (Logistic Regression, Genetic Algorithm, Back-Propagation Neural Networkand Genetic Algorithm Back Propagation Network) are compared, to indicate which one is close related to the real market, respectively. Based on simulation results, the method of Back-Propagation Neural Network is obviously better than the one of traditional Logistic Regression;To pick variables, genetic algorithm is superior to stepwise regression; Furthermore, if we adopt the genetic algorithm to select the network architecture, it will help to improve forecasting performance of BPN; Regardless of the research methods, the more reference variables are considered, the better the performance will be improved in this study.