在波幅震盪股市裡，公司無預警倒閉下市造成投資人嚴重損失，是現今公司在 企業經營、決策管理及個人理財應重視的課題。本研究透過電腦萃取有效上市公司 的財務資料，運用財務變數準確預測股價報酬率，從雜亂無章的交易資料中抽絲剝 繭找出蛛絲馬跡，轉成有用知識，進而作出正確的決策，能協助公司及投資人規避 風險提高獲利空間。本研究經由台灣股票上市公司選擇六個不同產業的財務報表， 資料收集自台灣經濟新報金融財經網站TEJ資料庫，財報中挑選條件屬性25個和1個 決策屬性-股價報酬率。運用屬性選擇建立模型和不同分類器評估，如堆疊集成 (Stacking) 、簡易貝氏分類器 (Naive Bayes) 及徑向相基底函數類神經網路 (Radial-Basis-Function Network)三項評估績效。找出影響股價報酬率之重要決策因 子，提高預測的正確度。實證結果為流動負債、利息保障倍數、每股營業利益為影 響股價報酬率之重要決策因子；再者，對財務資料言，使用分類器不同則得到不同 的績效衡量。

In the volatility of the stock market, the company did not return to the market caused by a serious loss of investors, is the company in the business, decision-making management and personal finance should pay attention to the subject. This study uses the computer to extract the financial information of the effective listed companies, use the financial variables to accurately predict the return rate of stock price from the chaos of the transaction information in the spinning to find clues into useful knowledge, and then make the right decisions to help companies and investors to avoid risk raising profit margins. In this research, we study the data from the Taiwan’s listed companies in six different industries of financial statements, based on the Taiwan Economic Journl (TEJ) online financial database. In the financial statement data, 25 condition attributes and one decision attribute –return rate of stock price - are selected. Using attribute selection to establish models and different classifier evaluations, such as Stacked Integration (Stacking), Naive Bayes (NB) and Radial-Basis-Function Network (RBF Network) three assessment performance. Find out the important decision factors that affect the return rate of stock price, and improve the correctness of the forecast. The empirical results are current liabilities, interest protection multiple and earnings per share business interestare important determinants that affect the return rate of stock price. From the empirical results, use different classifiers for financial data has different performance measures.