| Summary: | 碩士 === 國立中興大學 === 應用經濟學系所 === 101 === The objective of this research is to compare the application of ARIMA and HAR stock analysis models and their predictive abilities. This research selected two companies for each of the three stock price scenarios (price rising, price decreasing, and price consolidation) as empirical research objects. The effectiveness of ARIMA and HAR in predictings stock prices were tested by statistical indicators and simulated investment.
In this study, ARIMA outperformed HAR in terms of RMSE, MAE, MAPE and PMSPE only in the two companies whose stock prices are on the increase. In the situation of the other four companies, the predicted results of statistical indicators show no major differences when the ARIMA or HAR is applied. In terms of Lewis modeling capabilities, ARIMA and HAR both demonstrate outstanding accuracy. In the aspect of simulating investment, the ARIMA model of two companies outperforms the others in investment profits; however, the other four companies perform better in HAR model. Furthermore, from the perspective of trend, the investment profits of long-term held are better than the investment profits of two companies whose stock prices are consolidative in all estimated model. In the trend of increasing stock prices, the best investment profits in all estimating models are almost equivalent to the investment profits of long-term held stocks. In the trend of decreasing stock prices, the best investment profits in all estimating models are obviously better than investment profits of long-term held stocks. Therefore, the result explains that both ARIMA and HAR possess remarkable predicted capabilities, which can be the references for those who want to research HAR model.
|
|---|
