運用類神經網路學習公司最常被研究報告引用之財務資料並分析其與公司股價行為關連現象

Abstract

本研究的目的主要在於運用類神經網路學習過去經常被研究報告引用之財務基本面資料並分析在不同數量之財務資料種類其與股價關連程度的分析。首先經由大量文獻整理，以被引用頻率的高低挑選出最常被各文獻所使用之財務指標１０個，進而運用類神經網路試誤法進行財務指標之逐步篩選，挑選出最影響股價指數之財務指標種類。 研究標的為民國八十四年一月至八十八年五月之電子類股指數與電子類財務指標，為探究關鍵財務指標預測未來某月之指數較具影響力?本研究以財務指標公佈之本月，下一月，下二月及下三月的電子類股指數為輸出進行驗證。 經由實證結果，所挑選之關鍵財務指標順序如下：速動比率、純益率、總資產報酬成長率、負債比、流動比率、總資產週轉率、固定資產週轉率、存貨週轉率、應收帳款週轉率、資產報酬率。當運用所選出關鍵財務指標預測未來電子指數時，依序輸入關鍵財務指標達3個以上，進行股價指數之預測，預測之誤差率便可降低。 而運用財務指標於預測未來電子類股指數，於預測未來股價指數時，以預測當月之電子類股指數可得到較小的絕對值誤差百分比（MAPE），顯示利用財務指標預測未來股價指數時，以預測當月之股價指數可獲得較好的結果。故可推演出股價似有領先反應基本面報表產出時間，顯示台灣市場似為非健全市場，公司之基本面資料在尚未公告前即已率先反應，似有先知先覺者在基本面資料尚未公告前即得知公司成長與否。

The study is to analyze the relationship between the company’s financial statements and it’s corresponding stock index. On the basis of lecture survey, ten financial factors are collected initially. And through the methodology of try and error in neural network simulations some critical financial factors are decided and selected for related use step by step. From the pre-experiment, the critical financial factor are collected and ordered as follow: acid test ratio, profit margin on sales, return on total assets ratio, total debt to total assets, current ratio, total assets turnover ratio, fixed assets turnover ratio, inventory turnover ratio, average collection period, return on total assets. The study is shown that when the critical financial factors are decided and when the number of critical financial factors are up to 3, the mean absolute percentage error (MAPE) of the electric stock index can be reduced to 7%, and the average MAPE is 11%. The study is also shown that when using critical financial factors to predict the following months stock index, surprisingly, the current month stock index always get a better forecasting result compare with the months of next. And there is a conclusion that the critical financial factors do have relationship with stock price. Furthermore, the stock index do have reflected the fundamental factors before the news was disclosed. It seems to show that there are some diviners who have known the fundamental information of a company before the news is published in Taiwan stock market.