| Summary: | 碩士 === 國立成功大學 === 統計學系碩博士班 === 96 === At present, the globalization and the development of the BRICs lead the global steel and iron ore demand to increase. In addition, with the increasing oil price, the original material price have successive years of risen. That proves the inflaction impact is inevitably. However, The bank account interest rate of Taiwan is low, and the economy development is also hampered by state policy. If people want to gain additional wealth in this hardship environment, finances investment will be the only way. According to statistical data derived from Taiwan Stock Exchange Corporation, the stock market is the main investment pipeline for the domestic investor. By above explanation, this research focuses on the forecasting of the steel and iron industry stock price index.
This research data includes technical variables, macro-economical variables, international stock variables and Taiwan stock informations. The research scope is from Jan.,2000 to Dec.,2006. We establish forecasting models by Regression Analysis、Time Series Analysis（ARIMA）, Back-Propagation Network（BPN）and Adaptive Network-Based Fuzzy Inference System（ANFIS）, and compare accuracy of all the models. The result showed that Time Series is the best forecasting model, and next for Regression Analysis. ANFIS is strong to train sample, but it is the worst to forecasting. BPN get a better performance than ANFIS as a whole, but is inferior to the statistical methods.
|
|---|
