股價泡沫預測與投資人行為偏誤:粗集理論之運用

Stock Bubble Prediction and Investor Behavior Bias: Application of Rough Set Theory

泡沫在人類的經濟發展史上始終存在，因此，泡沫的形成原因，一直是學者專家們研究的焦點，台灣股市30多年來歷經多次的股價泡沫，本研究從投資人的偏誤行為角度採用粗集理論進行分析。研究期間為1988年7月至2019年12月，採計動能、從眾、反從眾、投資人情緒、週轉率、融資餘額、貨幣供給與領先指標等9項變數。實證結果發現，當以9種變數的訓練樣本為基礎，第13條決策規則，即貨幣供給和融資餘額屬性數同時為1，預測泡沫期間的準確性最高；當以測試樣本為基礎，發現股價泡沫預測正確率為90％；在進一步的研究中，以4種約簡屬性變數為基礎，第7條決策規則，即動能、融資餘額與週轉率屬性數同時為1時有較高的準確性；當再以測試樣本為基礎，發現股價泡沫預測正確率為82.2％。研究結果對判斷目前股價是否泡沫，有很高的參考性。

Bubbles have always existed in the history of human economic development. Therefore, the reasons for the formation of bubbles have always been the focus of research by scholars and experts. Taiwan's stock market has experienced multiple stock price bubbles for more than 30 years. This study adopts a rough approach from the perspective of investor bias Set theory for analysis. The study period was from July 1988 to December 2019, and 9 variables including kinetic energy, herd, anti-herd, investor sentiment, turnover rate, financing balance, money supply and leading indicators were adopted. The empirical results found that when the training samples of 9 variables are used as the basis, the 13th decision rule is found, that is, the number of attributes of money supply and financing balance is 1, and the accuracy of predicting the bubble period is the highest; The correct rate of stock price bubble prediction is 90%; in further research, based on 4 reduced attribute variables, it is found that the seventh decision rule, that is, when the number of attributes of kinetic energy, financing balance and turnover rate is 1, there is a higher Accuracy: When based on test samples, it is found that the correct rate of stock price bubble prediction is 82.2%. The research results have a high reference value for judging whether the current stock price is bubble.