Two-Stage Least Squares based Iterative Identification Algorithm for Box-Jenkins Model

Abstract

In order to improve the parameters estimation precision, a two-stage least squares iterative algorithm for Box-Jenkins models is presented, which is based on the interactive estimation theory of the hiera rchical identification and the auxiliary model. The main idea of the algorithm is to decompose a Box-Jenkins system into two subsystems so as to identify each subsystem, respectively. In our algorithm, the dimensions of the involved covariance matrices in each subsystem turn to be small. The simulation results indicate that the proposed algorithm is effective and has a high computational efficien cy.