摘要
国库现金流预测对于提高国库现金的精细化管理水平具有重要意义。实证结果表明,相比指数平滑法、ARMA模型等传统预测方法,混频数据模型综合利用高频数据(日度数据)和低频数据(月度数据)对国库现金流的预测精度有所改善。MIDAS混频数据模型没有全面涵盖长期性、周期性因素,如与含长期性、周期性因素的线性回归模型、灰色系统模型、ARIMA模型等相互结合,有利于进一步改善预测精度。
The treasury cash flow forecast is of great significance for the fine management to improve the level ofthe state treasury cash. The empirical results show that,compared with the exponential smoothing method,ARMAmodel and other traditional prediction methods,the high frequency data of comprehensive utilization of mixing datamodel(daily data)and the low frequency data(monthly data)improves on prediction accuracy of treasurycash flow. MIDAS model,which has no comprehensive coverage of long-term,seasonal factors,will further improvethe prediction accuracy combined with the Linearing regression model,Gray system model,the ARIMA model.
出处
《区域金融研究》
2015年第2期43-45,共3页
Journal of Regional Financial Research