Forecasting Stock Index Using Diversified Models into Taiwan 50 Index

• Main Authors: Hsu,Wen-Pi, 許雯筆
• Other Authors: Lin,Ping-Chen
• Format: Others
• Language: zh-TW
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/35462536586263120110

碩士 === 國立高雄應用科技大學 === 金融系金融資訊碩士班 === 104 === It is a typical issue on forecasting the trend of share market. In the past, two technical systems are most used on forecasting share market, one is statistical models the other is and artificial intelligence model. On this research, we combine two technical systems and use together. The method to take 3 variable, None: Unselective Variable, PCA: Principal Component Analysis, SR : Stepwise Regression Analysis. 4 forecasting models are BPNN：Back Propagation Neural Network Model, MR: Multiple Regression Model, ES: Exponential Smoothing, and ARIMA Model. Finally, forming 3 forecasting sets to find out which is more accurate model on 4 forecasting models which are assessed on Taiwan stock index 50.According to outcome of research, it is better on the model by forecasting through selective variables. It is little different on comparing forecasting models and MR is better than BPNN. On this basement, we can find out accurate forecasting model through cross validation among many kinds of forecasting sets.