摘要
现实中的决策大多建立在对数据的预测之上,但研究者通常很少涉及预测的不可靠性问题,而暗含假定预测是可靠的。基于目前预测残差为正态白噪声的预设,提出了预测可靠性的概念,以及预测可靠性的度量方法,同时以需求预测中常见的ARIMA(p,1,q)过程MMSE(最小均方差)预测为背景,分析了可靠性与预测精度的关系,以及模型参数对可靠性的影响,初步讨论了可靠性在预测期决策方面的作用。
Many real decision-making problems are modeled and solved on forecasting data. Despite the fact that almost all forecasts are unreliable, researchers rarely address it. This paper proposes the concept of forecast reliability, and new measurement for forecast reliability. Based on Autoregressive Integrated Moving Average (ARIMA) process and its Minimum Mean Square Error (MMSE) forecast, the paper compares forecast accuracy with forecast reliability, and analyzes the relationship between specific process parameters and reliability, and discusses the influence of reliability on decisionmaking on forecast horizon.
出处
《预测》
CSSCI
2007年第4期59-63,80,共6页
Forecasting
关键词
预测
预测可靠性
ARIMA
预测期
demand forecast
forecast reliability
ARIMA
forecast horizon