摘要
基于时变Hurst指数提出一种用来描述时间序列长记忆性的具有时变分数差分算子的ARFIMA模型(时变分整自回归滑动平均模型),并给出建模的步骤。然后对新台币汇率时间序列建立时变ARFIMA模型。将建立的时变ARFIMA模型与其它模型进行了比较研究。证明了时变ARFIMA模型与其他模型相比较的优效性。
This paper develops a new ARFIMA(Auto Regressive Fractionally Integrated Moving Average)model-Time-Varying ARFIMA,with time-varying fractional differencing parameters based on the time-varying Hurst exponents and gives an estimation method.This paper builds a time-varying ARFIMA model of the time series of Taiwan Dollar.An empirical comparison with other models indicates that the Time-Varying ARFIMA model is more successful at presenting the currency exchange rate time series.
出处
《计算机工程与应用》
CSCD
北大核心
2004年第36期205-207,共3页
Computer Engineering and Applications
