本研究認為財務績效及財務狀況的優劣及個別公司差異與股價波動劇烈程度相關,因此以股價標準差為股價波動劇烈程度之代理變數,並將股價標準差由低到高分為四個級別,分別代表股價波動平穩至劇烈的程度,並以財務指標等變數利用決策樹方法及徑向基底類神經網路方法進行股價波動劇烈程度之預測,發現以財務指標預測整體樣本股價波動劇烈程度分類,各不同波動劇烈程度級別平均預測精確率最高可達到56%,對於股價波動最小之級別預測能力更為突出,其預測精確率更高達73%,而股價波動最劇烈級別預測準確率約為53%,研究發現股價波動平穩之樣本,其財務指標特徵明顯,但股價波動劇烈之財務指標特徵較薄弱,本研究經由加入股價淨值比、前期股價標準差等變數後改善股價波動較高級別之預測能力,本文模型預測下季度股價波動劇烈程度將小於樣本平均股價波動劇烈程度之精確率達97%,表示投資人可由本文預測波動最小級別之樣本中挑選股票將有97%的機率下季股價波動劇烈程度小於樣本平均值,另外本文預測下季度股價波動大於樣本平均之精確率達94%,其預測結果足以用於投資策略。
This study suggests that the company's financial situation will affect stock price volatility. Therefore, we use the standard deviation of stock price as dummy for stock price volatility. First, we divided standard deviation of stock prices into four levels high to low, representing the degree of its volatility from smooth to violent. Second, we take financial ratios to predict its volatility with a total of three methods as follow: (1) the number of C 4.5 decision tree method (2) Random Forest decision tree method and (3) Radial Basis Function Neural Network(RBFNN) method. The study found the average prediction accuracy rate is up to 56%. Stock price volatility of the lowest level of the prediction reaches 73%. And the highest level of stock price volatility prediction accuracy rate is also up to 53%. The study found stock price volatility of the stable samples show the obvious characteristics on financial indicators, but the serious samples of stock price volatility indicate weaker relationship on their financial indicators. Our study try to improve the ability of prediction of the high level by increasing price-to –book ration and the prior period stock price volatility. Our model forecasted the sample of stock price volatility will be less than the average stock price volatility in the following quarter and its accuracy is up to 97%.Besides, Our result shows stock price volatility is more than the mean of samples, which explanation of prediction reaches 94%. The results of this study can be used for investment decisions in practice.