Building Vector Autoregressive Models Using COMBI GMDH with Recurrent-and-Parallel Computations

摘要

The paper presents theoretical grounds of recurrent-and-parallel computing applying for modelling and prediction of complex multidimensional interrelated processes in the class of vector autoregressive models. The combinatorial GMDH algorithm is used for vector autoregressive (VAR) modelling by exhaustive search of all possible variants and finding the best model for every time series. The algorithm with selection a few best models for every process is used. The procedure of structural and parametric identification of VAR models is proposed. It is applied the combining all possible variants of systems from the selected best models and choosing the best system model according to an additional criterion. The test experiment on solving the problem of structural and parametric identification for experimental testing of algorithm efficiency was carried out. The effectiveness of constructed algorithm is demonstrated by prediction of interrelated processes in the field of Ukraine energy sphere with the purpose of effective managerial decision making.