Forecasting Long Memory Time Series for Stock Price with Autoregressive Fractionally Integrated Moving Average

摘要

The presence of long memory time series is characterized by autocorrelation function which decrease slowly or hyperbolic. The best suited model for this time series phenomenon is Autoregressive Fractionally Integrated Moving Average (ARFIMA) that can be used to model historical stock price in financial data analysis. This research is aimed to assess the ARFIMA modeling on long memory process with parameter estimation method of Geweke and Porter Hudak (GPH), and applied to opening price of Kedaung Indah Can Tbk Stock from May 2nd 2005 until March 26th 2012. The best suited model is found ARFIMA(5,0.452,4) where for short time forcasting is shown very close to actual stock price with small standard error.