A time series is split into two parts by using Singular Value Decomposition (SVD): one part reflecting the regular effects of exogeneous variables, the other reflecting random effects. The problem of variability of the end points of the constituent series in SVD is resolved. A method is developed of distinguishing the regular part of the series from the random part. The regular part is forecast using ARMA while the random part is forecast based upon a Fokker–Planck equation. This approach is compared to that of a full ARMA forecast on the whole original series.

中文翻译：

---

通过使用奇异值分解拆分时间序列，然后使用ARMA和Fokker Planck方程进行预测

使用奇异值分解（SVD）将时间序列分为两部分：一个部分反映了外生变量的规则影响，另一部分反映了随机影响。解决了SVD中组成序列的端点可变性的问题。开发了一种将序列的常规部分与随机部分区分开的方法。常规部分使用ARMA进行预测，而随机部分基于Fokker-Planck方程进行预测。将该方法与整个原始系列的完整ARMA预测方法进行了比较。