標題: 應用總體經濟因素於加權股價指數的預測-類神經網路與多元迴歸比較之研究
On the Prediction of Taiwan Stock Index with Macroeconomic Factors ─ A Comparative Study of Artificial Neural Network and Multiple Regression
作者: 陳瑞卿
Jui-Ching Chen
陳安斌
資訊管理研究所
關鍵字: 倒傳遞類神經網路;總體經濟因素;股價指數;落後期;back propagation network;macroeconomic factors;stock index;time lag
公開日期: 1999
摘要: 股票市場是上市公司籌措資金的地方,現也漸漸成為國人投資的主要管道。雖因短期供需未達平衡而使得投機盛行,但以長遠分析,影響股價指數趨勢的因素仍為總體經濟等基本因素所決定。但總體經濟的因素非常多,實際上與股價指數有較大關係的這些總體經濟因素與股價指數的關係究竟為何,以及在不同的時間下其所影響股價指數的程度是本論文所欲探討的。本論文嘗試建立一以類神經網路為基礎的股價指數預測模型,並與多元迴歸做比較。經實證研究可規納二點主要的結論:1、類神經網路於股價指數的預測優於多元迴歸模型。2、經過適當的資料前處理及了解各變數與股價指數的領先落後關係,類神經網路其預測平均誤差可有效的降低。另本研究也發現在考慮落後關係的情況之下對預測也有較好的結果,有助於投資人作出較好的投資決策。
The stock market is the major place of the listed company to raise funds, as well as the main channel for people to invest. Although unbalance of demand and supply makes gamble prevailing in short-term . In long-term analysis however, it is still the macroeconomic and other fundamental factors to decide the influence to Taiwan Stock Index trend. However, fundamental factors are too many. So, the problems are what macroeconomic factors are really more related to stock index. To find the relationship and the degree of influence to stock index in different time is important. Thus, this research is tried to build a neural network based stock index prediction model and compared the efficiency with multiple regression model . Two major findings of empirical study in this research are as following: 1. the forecasting results from the neural network techniques seems better than that from the multiple regression model. 2.Through the data pre-processing of adjusting the most fitted time lag of each factors, the mean error can be reduced significantly. It’s benefit to the investor to do better decisions.
URI: http://140.113.39.130/cdrfb3/record/nctu/#NT880396017
http://hdl.handle.net/11536/65597
Appears in Collections:Thesis