摘要
首先研究上证指数日内高频成交量时间序列的统计特征,包括平稳性、自相关和长记忆性,然后我们通过对剔除日内周期趋势的成交量序列建立ARMA模型,并分别结合ARCH类模型和ARFIMA模型消除模型的异方差和长记忆性.我们的实证分析结果表明在消除日内周期项、异方差和长记忆性后建立的时间序列模型比原始序列的时间序列模型有更高的预测精度.
Statistical feature of intraday high-frequency volume time series, including stationarity, autocorrelation and long-memory feature were analyzed. The mean-variance models for intraday volume time series were established which have removed intraday periodical w-shaped trend. By setting up ARCH-type model and ARFIMA model to eliminate Heteroscedasticity and long-memory feature, the outcome of simulation and forecast were improved in a large degree with ARFIMA model.
出处
《北京师范大学学报（自然科学版）》
CAS
CSCD
北大核心
2013年第1期85-89,共5页
Journal of Beijing Normal University(Natural Science)
基金
国家自然科学基金资助项目(61174165)
