B1546
Title: High-dimensional functional time series prediction model solved with a mixed integer optimization method
Authors: Nazgul Zakiyeva - Technische Universitat Berlin (Germany) [presenting]
Abstract: A network functional autoregressive model is studied for large-scale network time series. The estimation of the proposed model is approached using a mixed integer optimisation method. By including the high-dimensional curves, the proposed model captures both serial and cross-sectional dependence in the functional time series network. The methodology is illustrated on large-scale natural gas network data. The model provides more accurate several days-ahead hourly out-of-sample forecasts of the gas in- and out-flows compared to alternative prediction models.
