標題: 以自組織映射圖結合倒傳遞網路模型學習財務報表資料預測公司股價
Analyzing the Financial Statement Using Self-Organizing Map Combine Back-Propagation Neural Networks to Forecast Model
作者: 李宗昭
Tseng-chao Lee
陳安斌
An-Pin chen
管理學院資訊管理學程
關鍵字: 類神經網路;自組織映射圖;倒傳遞網路;財務比率;財務報表;Neural Networks;Self-Organizing Map;Back-Propagation Neural Networks;Financial Ratio;Financial Statement
公開日期: 2005
摘要: 隨著資訊科技的進步,電腦的速度與功能也愈來愈強大,更因為電腦的運算速度的提升,使得人工智慧技術也漸漸地成為協助處理複雜難題的工具。而類神經網路是近年來在人工智慧技術領域中相當重要的一種方法,且各項發表與實證指出,類神經網路在分類與預測公司股權價值上有相當良好的參考性。故本實驗嘗試提出整合類神經網路用於分類的自組織映射圖(Self-Organizing Map;SOM)與結合預測的倒傳遞網路(Back-Propagation Neural Networks;BPN)的混合式類神經預測模型,並與單純倒傳遞網路(Back-Propagation Neural Networks;BPN)模型進行股價預測之比較與驗證。 研究標的為印刷電路板(Printed Circuit Board;PCB)產業公司為研究對象共十家公司,由該公司上市櫃時間次年到民國94年底,其財務資訊與股價,為研究變數與樣本,而研究的變數為引用相關整理於期刊論文中具影響性的財務比率,此可由財務報表中分別歸納為償債能力分析、經營能力分析、財務結構分析、現金流量分析、成長力分析的五種財務比率分類。本實驗共挑選具影響力的20個財務比率變數,以進行月均股價預測與模型比較之測試。 □ 研究顯示,對於財務指標用於預測每季財務報表的下一個月的月平均股價,均可得較小的平均絕對誤差(MSE)。預測結果也驗證,台灣股票市場屬於效率市場學說的半強勢效率市場理論,對於公司的財務報表屬基本分析的範疇內,其資訊在未公開前對股價具有影響性。
With the advance of information technology, the processing power and speed of computer increase dramatically. Such improvement also allows AI technology a feasible tool to assist on dealing with complex problem or scenario. In recent year, neural networks has become an important methodology in AI technology field. It is capable of producing good referential materials for categorizing and predicting companys stock rights worth. As a result, neural networks to categorized Self-Organizing Map and combine with predicted Back-Propagation Neural Networks to produce Hybrid-Neural Networks Forecast Model is applied. This study also compare and check Hybrid-Neural Networks Forecast Model with simple Back-Propagation Neural Networks model. The studying subjects of this research are ten companies in Printed Circuit Board industry. Financial information and stock price of each company to be used are collected as samples and variables. Significant financial ratio relating to this research is cited from journals and thesis of this field. The financial report can be categorized into five classification. These are credit standing analysis, operating ratio analysis, financial structure analysis, cash flowing analysis and growth analysis. Finally, in this expectant, it carry out middle month price predicting and model comparison. This research demonstrated that applying financial ratio in predicting the following middle month price after each quarterly financial report will generate smaller MSE. It shows the Taiwan stock market is closer to semi-strong form efficient market and proves un-published company financial report will affect the share price.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT009364529
http://hdl.handle.net/11536/80016
顯示於類別:畢業論文