Forecasting stock indices with relaxed variable kernel density estimation

• Main Authors: Jing-CyuanLin, 林靖荃
• Other Authors: Tien-Hao Chang
• Format: Others
• Language: zh-TW
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/89383861190762104646

碩士 === 國立成功大學 === 電機工程學系 === 102 === Stock prices is time series. Stock prices are basically dynamic, non-linear, and chaotic in stock markets. Stock markets are influenced by many factors. Predicting stock price or stock index with the noisy data directly is usually subject to large errors. In this research, we try to predict the monthly closing price data with Shanghai Composite Index, Standard ＆ Poor's 500 Index and Nikkei 225 Index from January 1993 to December 2009. We discover that while the tendency of the stock index price is obvious, turning the Shanghai Composite Index price of the predicted target into percentage change in Shanghai Composite Index price has the better prediction. We also do robustness test. We forecast S＆P 500 Index closing price and Nikkei 225 Index closing price. We also find that the prediction of using the RVKDE (relaxed variable kernel density estimator) has the better accuracy than SVM (support vector machine) and neural network. While adding the percentage change of the stock index of the three main export countries for target country, the prediction is also improved.