| 摘要: | The basic characteristics of stock prices are sensitiveness, non-stationarity and asymmetric volatility. Investors would like to forecast the movement of stock prices even though it is a stochastic process. Stock prices can be affected by a wide variety of macroeconomic, industry, and company-specific factors, including global stock price indices, aggregate economic activity, exchange rate, interest rate, among others. An abundance of researches can be found in literature, however, have made forecasts of stock prices by using their own lagged terms. This study applies the basic methods that could be applied more easily by investors than other mathematical methods including fuzzy time series model, wavelet analysis, neural network forecasting model, VAR, VECM etc. However, the empirical results are satisfied and interesting.
This study aims to forecast the movement of the Vietnam index (VN-Index) by exponential smoothing, Holt’s exponential smoothing, Winter’s exponential, empirical mode decomposition, and autoregressive integrated moving average (ARIMA) models, and generalized autoregressive conditional heteroskedasticity (ARIMA-GARCH) models. The data of VN-Index is collected at three kinds of frequency. They are the daily interval, which covers periods from November 2012 to October 2014 with 496 closing prices, the weekly interval, which covers the periods from November 2009 to October 2014 with 240 closing prices and the monthly interval, which covers the periods from July 2000 to October 2014 with 173 closing prices. The prediction accuracy of VN-index is compared among the methods in terms of two criteria, namely root mean square error (RMSE)and mean absolute percentage error (MAPE) based on out-of-sample forecasts of 15 days, 30 days, 60 days; 6 weeks, 9 weeks, 12 weeks; and 3 months, 6 months, 12 months.
Evidence reveals that whatever the daily, weekly or monthly stock price forecast, the ARIMA-GARCH approach is consistently ranked the first among the methods in prediction accuracy no matter what criteria are applied and what out-of-sample forecasts are based. This finding is of great significance to the investors in the Vietnamese stock market when they attempt to forecast the movement of the market. The use of ARIMA approach and the use of GARCH approach are combined to utilize power of prediction in case of the data have trend, heteroskedasticity as natural stock returns. This study also proposes the advantaged hybrid model to forecast monthly stock returns. Exponential smoothing families predict the movement of stock prices is very well. However, there are a few of methods, which have great empirical results. |
