應用基因法則最佳化類神經網路建立財務分析模式之研究

Abstract

本研究以類神經網路為基礎，結合同樣源自生物科學，普遍用作最佳佳化 工具的基因法則，建立一最佳化類神經網路財務分析模式，對台灣股票市 場進行基本面的財務報表分析。期望在類神經網路的最佳化下，徹底的揭 露財務報表所隱藏的重要資訊。本研究所提出的財務分析模式是根據上市 公司的財務報表，建立財務結構、償債能力、經營能力、獲利能力、成長 能力等五個財務項目的評分表。然後經由最佳化類神經網路的學習，掌握 評分表內的知識規則，繼而將上市公司實際的財務指標值輸入類神經網路 ，以得到各上市公司在不同財務項目上的得分，並綜合各財務項目的得分 ，評估上市公司財務體質的良窳。最後，進一步利用財務分析的結果預測 下一年度公司的每股盈餘。本研究並以電子業的十家上市公司為研究的對 象，以實際的財務資料說明如何以基因法則最佳化類神經網路進行財務分 析。而由研究的結果，則驗證了本研究所發展的財務分析模式的可行性。 The Study is applied the genetic optimal neural networks to develop financial analysis model based on the company financial statements in Taiwan Stock Market. It is expected to completely disclose the important information hid in financial statements by combining the power of genetic algorithms and neural network. According to the financial statements, the financial analy- sis model proposed by the study constructs five scoring tables of financial items related to capital structure, liquidity, operating perforemance, return on investment and growth analysis respectively. After the knowledge-rules shown in scoring tables through trianing optimal neural networks being entirely captured , the data of financial ratio collected in this study is input to obtain each company's score in each financial item. Finally, each company's score is used to evaluate company financial health and forecast the earning per share next year. Ten electronic companies in Taiwan Stock Market are selected to study as subjects. And the actual financial data of those ten companies is utilized by the genetic optimal neural networks to explain how to perform financial analysis. From the results of this case study, the financial analysis model proposed by the study is verified which is feasible to evaluate company financial health.