| Summary: | 碩士 === 國立嘉義大學 === 資訊管理學系研究所 === 107 === This study compared the traditional time series and fuzzy time series prediction of Taiwan's stock market performance. The stock picking methods were adopted by nine world-renowned investment experts. The better stock portfolio was calculated with dynamic programming in Knapsack problem and genetic algorithm. We used moving average, Stochastic Oscillator (KD), Moving Average Convergence/Divergence(MACD) and Bollinger Bands(BBands) as four trading strategies to form 120 investment portfolios. Finally, we compared the return rates of each portfolio for performance evaluation. In the study, approximately 635 Taiwan-listed companies, a total of 1,794,510 stock price data during January 2, 2008 to October 31, 2018, were used. The stock prices for the previous months were used as training data to predict the stock price after 5 days. The results suggested that fuzzy time series is better than traditional time series. Useing Michael Murphy's stock picking method in 120 portfolios, the stock portfolio with the Knapsack problem and the MACD trading strategy achieved the highest return rate of 36%. On the other hand, the worst investment portfolio was Peter Lynch's stock picking method with random stock portfolio and BBands trading strategy, return rate only -16%.
|
|---|
