Self-Organizing Maps Combined with Neural Networks are Applied to Modeling and Prediction for Taiwan-Weighted Stock Index

• Main Authors: LIAO,YI-MING, 廖翊銘
• Other Authors: TIEN,HUI-CHUN
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/dv752a

碩士 === 靜宜大學 === 財務與計算數學系 === 105 === Recently, AlphaGo has demonstrated the great potential for artificial intelligence. The development and application of machine learning and Neural Network also obtain a lot of attention. These two techniques are used to retrieve valuable information from data and then to predict important futures. In this study, we use Self-Organization Map(SOM), which is developed by Kohonen(1982) to select presentative Taiwan-Weighted Stock index Technical Indicators. The principle of SOM is to make data clustering from, and then take representative Technical Indicators from each clustering as an input unit of Back-propagation Neural Network(BPN), so that input units are basically with different characteristics. Using daily trading information of Taiwan-Weighted Stock to calculate each input Technical Indicators values, and then theses indicator values are feed to our training model, model output values are the consecutive trading day close Stock index. Open source R is used in this study for data analysis as well as SOM and BPN modeling.