| Summary: | 碩士 === 國立臺灣科技大學 === 資訊管理系 === 95 === Because of the high return rate, stock is always one of the popular investment targets. Considering the high level of risk based on high return rate, investors need more information and investment strategies to make a decision when trading stock.
Technical analysis is an important tool which depends on the past stock price, volume data and different charts to predict the fluctuation of stock market. In order to confirm the effectiveness of technical indexes, we select some technical rules to predict the trend on stock market.
In our research, we utilize genetic algorithm to decide the combination of technical rules and their weights. Then we proposed a trading system, which can choose different trading strategies to simulate the stock that investors buy or sell. IBM, Intel and Wal-Mart are investment targets within Dow Jones Industrial Index and the testing period is between 2001 and 2004.
As the result, we found that in spite of the results of Wal-Mart our experiments can always beat the buy and hold benchmark method. In addition, using multiple transaction trading mechanism had a lower risk than single transaction trading mechanism and signal collision cancellation mechanism has better return than signal collision recognition mechanism. We also tested the different lengths of learning period and deduced the robust result when setting the length as one year and one month. Finally, our results after concerning transaction cost still exceeds the return rate of buy and hold method between 17% and 62%. Hence it is believed that our research provided a good strategy method in stock trading.
|
|---|
