摘要

Closed solutions to the Fokker-Planck equation with linear and quadratic diffusion terms have been generated. These are applied to forecasting time series to generate a probability distribution for the next point, rather than a single point estimate as in autoregression. The parameters in these models are dynamically estimated from past data. The resulting solutions are compared to a conventional autoregressive approach with encouraging results.

  • 出版日期2015-2-15

全文