A subspace identification algorithm for the Wiener model is reformulated from a control point of view. The linear time-invariant part and the inverse static nonlinear function are simultaneously identified without an iterative procedure, and then the Wiener input/output data-based predictive controller is designed on the basis of the identified model. Through simulation and experimental studies for the multivariate control of polymer properties in a continuous methyl methacrylate (MMA) polymerization reactor, the proposed identification method is validated, and the performance of the designed predictive controller is examined. The Wiener model identified without any iterative procedure is found to predict accurately the output of the polymerization reactor, and the predictive controller designed in this work performs quite satisfactorily for polymer property control in the polymerization reactor.