SCSC2003 Abstract S8566

Periodic Time Series Evaluation & Model Construction

Periodic Time Series Evaluation & Model Construction

Submitting Author: Dr. Jerry Couretas

Abstract:
Cycles, common to all continuous time phenomena, are also key elements of both stationary and non-stationary time series descriptions. Using this insight, a method is developed for decomposing a discrete time series into its period-based elements, each of which is then a candidate representation for the time series of interest. Correlating to the original output description determines each cycle component’s applicability as a “partial model” description of the original time series. Projecting the relevant (correlation wise) cycles forward in time then produces a behavior forecast.


Back to SCSC2003 Abstracts