MS

M.W. Sibeijn

1 records found

In this thesis, I introduce a novel model identification approach to time series forecasting. For linear stationary processes, such as AR processes, the direction of time is independent of the model parameters. By combining theoretical principles of time-reversibility in time se ...