Detecção de mudança de nível em séries temporais não lineares usando Descritores de Hjorth

Abstract

The purpose of this paper is to present a method for detecting the dynamic changes in a nonlinear time series that uses the Hotelling T2 multivariate control chart to monitor the variation in three normalized descriptors: Hjorth's descriptors of activity, mobility and complexity. This approach was applied to different simulated nonlinear time series and is illustrated in this paper with a detailed example. A case study with six time series of short-term electricity load consumption was also used to confirm the method's effectiveness.