Title: Segmentation of autoregressive processes via the cross-entropy method
Authors: Lijing Ma - Macquarie University (Australia) [presenting]
Georgy Sofronov - Macquarie University (Australia)
David Bulger - Macquarie University (Australia)
Abstract: One approach to modelling nonstationary time series data is to use change point detection methods to optimally segment the signal into intervals within which the process behaves stationarily. We assume that each segment is an autoregressive time series with its own model parameters. Taking the nonstationarity into account and identifying the number and locations of these structural breaks are of interest. Our method includes two steps. The first step is to use a distributionally tailored cross-entropy method to identify these potential change points to segment the time series. Once these potential change points are obtained, modified parametric spectral discrimination tests are used to validate the proposed segments. A numerical study is conducted to demonstrate the performance of the proposed method across various scenarios and compare it against other contemporary techniques.