Summary: | 碩士 === 大葉大學 === 事業經營研究所 === 88 === ABSTRACT
The most concerned issue for participants in stock market is to predict the trend of stock price in advance. Since the equity of each financial institute in the stock market is very large, it’s vibration will affect the whole market significantly. Besides, since it’s future index has become a large trading objective, it becomes more concerned
recently.
Based on the reasons mentioned above, this research tries to us
ARIMA model to (1) analyze and forecast financial stocks index in Taiwan, (2) generate an analysis model to evaluate the accuracy of the ARIMA model developed.
This research use daily financial stock index and weekly financial stock index to do the time series forecast. Our results are as follows:
1. By using the daily stock index data as sample, the stock price seems to follow random walk pattern. By combining AR(1) model, and using average price for each days span, the forecasting error reduce to around 2.2%.
2. By using weekly stock index as sample, the assumption of random walk become invalid, and the best forecasting model is ARIMA(1,1,1), the average forecasting error is 4%.
3. The day-to-day forecasting will be affected by random vibration of stock price, the forecast accuracy is not accepted. However, if we try to forecast the stock price 3-5 operation days from now, and use the stock price for next few days to make certain adjustment, we can obtain a better result.
Key Words : ARIMA, finance and insurance stock price, time series
|