Improving Fuzzy Time Series Model Based on Price Fluctuation Limit Forecasting Taiwan Stock Prices

[[abstract]]在投資股票市場時，投資人時常須面臨「選擇投資的標的」、「決定買賣的時機」，這兩個問題，此時若有能有效預測股價、股票指數等資訊的輔助工具，便能以更加客觀、系統化的方式做出決策；本研究以漲跌幅限制為基礎，改良了由許多前人研究證實在預測歷年大學入學人數、氣溫、台灣加權股價指數上都有良好表現的模糊時間序列模型，以股價資料與漲跌幅之間的轉換，結合前人提出、廣受應用的數個模糊時間序列模型，使模糊時間序列模型能夠更充分得利用歷史資訊，並免除了前人易遇到、使預測準確度降低的出界問題；最後將提出的模型應用到台灣股票市場中，用以預測數檔具代表性的股票股價，其中包含了能充分反應台灣股票市場整體股價變動的股票─台灣50ETF，以及世人視為最能代表台灣股市整體漲跌狀況的台灣加權股票指數(TAIEX)，並與前人的研究相比較。[[abstract]]Investors in stock markets always need to handle portfolio selection and trading timing. We will be able to make more objective decisions in more systematic way if there is a tool which helps forecasting helpful information such as stock prices and index values while dealing with these problems. In this paper, we improved the Fuzzy Time Series basing on fluctuation limits in stock markets, while the Fuzzy Time Series has been proved having good performance in forecasting enrollments, temperature, and value of TAIEX by many former researches. Combining some popularly applied Fuzzy Time Series models with transforming between stock prices and price fluctuations, our proposed method enables these models to utilize information from historical data more completely. Even more, our proposed method can keep the quality of forecasting while there is possibility for actual price to exceed the upper bound or lower bound of the universe of discourse in conventional Fuzzy Time Series. We also applied the proposed method to forecast some representative stocks including Taiwan 50 ETF which could effectively reflect integrated fluctuations of prices of stocks in Taiwan stock market. We test the forecasting performance of the proposed method on TAIEX to compare its experimental results with former research as well.[[note]]碩