上市公司依據規定公告的財務報表資訊,提供投資者最直接獲取企業動態的最佳管道,然而一般投資者因為沒有足夠的判讀能力來了解財務報表所帶來的訊息,也因此沒有隨時更新個股的投資價值。本研究與現有文獻的不同點在於利用機器學習的方法,以台灣股市個股每季公告的財務報表財務比率變數,作為對個股收盤價的訓練資料,讓演算法找出規則,預測下一季公司價值,將預測的公司價值與收盤價作比較,再依投資策略進行股票作多或放空。實證結果發現,支持向量迴歸能有效預測個股投資價值,依據本研究投資策略買入具有投資價值的個股,持有期間一季、二季、四季,投資平均報酬率能打敗0050及發行量加權股價指數的平均報酬率。
Publicly traded company announces financial statement information based on the rules, providing the investors a straightforward channel to obtain enterprise trends. Whereas, general investors may not have adequate ability to interpret the messages delivered by financial statements, and thus could not update individual stocks’ investment value in time. The difference between this study and existing literatures is to apply the methodology of machine learning, using financial ratio variables from quarterly announced financial statement in Taiwan individual stock market as the training data for individual stock closing price, allowing the algorithm to figure out the rules to predict company values in next quarter and comparing predicted company value with closing price, then follows investment strategy to process stock long or short Sale. Empirical evidences showed, support vector regression can predict individual stock’s investment value effectively. Buying individual stock which has investment value based on the investment strategy in this research and holding one quarter, two quarters, four quarters, the average return rate of investment can beat the average return rate of 0050 and TAIEX.