| Summary: | 碩士 === 國立中山大學 === 財務管理學系研究所 === 106 === This paper attempts to maintain the originality and breadth of data. I have incorporated as much as possible of all market data (on a daily basis) related to the Taiwan Capitalization Weighted Stock Index (TWII), and have combined the macroeconomic data of Taiwan and U.S. (on a monthly basis). Then I use these data to build the Random Forest model to forecast the daily direction of TWII. There are 11 categories of the data used in this paper. After data conversion and parameter optimization, a total of 711 variables were obtained. All 19 Random Forest models were trained and forecast the daily direction of TWII by means of Walk-Forward Analysis (WFA), respectively. Then I use Stacking method to build another new Random Forest model, which uses all these prediction results to train. Then I will get the final prediction results of the direction of TWII. At the end, a simple investment strategy is designed by the final prediction results for investing in Taiwan Stock Price Index Futures. Then I can know the feasibility of investment by the model prediction results. The strategy profit factor was found to be 2.3, with an Annualized Rate of Return of 37.5% and a winning percentage of 65%.
|
|---|
