Abstract: An improved resource allocating network(IRAN)is employed in online modeling of nonli-(near) dynamic systems,which integrates the typical resource allocating network(RAN)algorithm,merging strategy for similar hidden units,deleting strategy for redundant hide units,and link weight learning with moving data window.IRAN can improve the accuracy and generalization of model effectively for nonlinear dynamic system modeling.The combined modeling method with IRBFNN and IRAN can be applied to adaptive modeling of nonlinear dynamic systems with uncertainties.Simulation studies show that the(model) built by the proposed method is of favorable performance on the parsimony,generalization and (adaptation) of model.