Predictive control is based on an intuitively appealing concept where the current control action is based on a prediction of the system controlled output at some time step in the future. Originated from chemical process engineering, several predictive control methods have emerged, including model algorithmic control (MAC), dynamic matrix control (DMC), extended prediction self-adaptive control (EPSAC), extended horizon adaptive control, multistep multivariable adaptive regulator (MUSMAR), and generalized predictive control (GPC). Sharing a similar philosophy, the details of these controllers are different from each other due to different choices of cost functions, constraints, and dynamic models. Almost all of them are model-based methods. A data-based treatment of predictive control will be presented, and it will be shown how it is an effective tool to suppress flow-induced vibrations. In this approach, the predictive controller gains are synthesized directly from input–output data instead of working through an intermediate identification model. Experimental results will be used to illustrate the benefit of this data-based treatment of predictive control over conventional model-based approach. A technique will be developed to avoid drifting of the controller gains identified with closed-loop data, and comparison with other conventional methods such as conditional updating, dither, and leakage will be made.