Cross validation for uncertain autoregressive model

Abstract

Uncertain time series models have been investigated to predict future values based on imprecise observations. The existing researches focus on how to estimate unknown parameters in the uncertain time series model without considering how to determine the lag order. This paper proposes three types of cross validation methods, i.e. fixed origin cross validation, rolling origin cross validation, and rolling window cross validation to choose the lag order considering the model’s prediction ability, and derives corresponding calculation methods under the framework of uncertainty theory. A numerical example and a real data example illustrate our methods in detail.

Topics & Concepts