Modeling strategy of high order ARMA model
Qi Wang, Li-Xin Wang, Qiang Shen · 2016
As a classical ARMA modeling strategy, time series analysis is used widely. But when the model order is high, the computational complexity will be very large. Dynamic data system can decrease the computational complexity by predigesting two-dimension order searching to one-dimension, but there are some problems, such as complexity of parameter estimate and artificial uncertainty. A modeling strategy of high order ARMA model is proposed, the main idea of which is using long auto-regression model to obtain the prior information of model order and avoid order searching, and it uses the model equivalence principle to predigest non-linear calculation to linear calculation in parameter estimating. The simulation results show that this modeling strategy decreases the modeling computational complexity greatly on the basis of ensuring modeling precision.