Most control engineers concentrate on finding a controller given the plant model or identifying a model from the data. There is no doubt that model-based control and system identification are closely related, simply because one depends strongly on the other. In this work a subspace identification algorithm for LF-LPV (linear-fractional linear parameter-varying) models is reformulated from a control point of view. This algorithm is referred to as an input/output data-based predictive control, in which an explicit model of the system to be controlled is not calculated at any point in the algorithm. It allows for the construction of a nonlinear model predictive controller for an unknown nonlinear system directly from a set of its open-loop measurements. As an example of the input/output data-based predictive control, the styrene solution polymerization in a continuous reactor system is considered to prove the superior performance of LF-LPV input/output data-based predictive controller for polymer. quality control. This approach gives a new angle for attacking the problem of identifying and controlling nonlinear systems.