The target of this article is to design a data-driven model predictive control algorithm for general multirate sampled-data systems. Multirate sampling widely exists in the industrial process control systems. In this paper, not only sampling periods between inputs and outputs are different, but also periods among inputs or outputs are different from each other. For such the general multirate sampled-data system, we combine the lifting technique and partial least square method to obtain inputs/outputs data sets, which are used for the model regression. An incorporating autoregressive exogenous (ARX) structure model is utilized to model predict. Then we give the principal components cost function for the model predictive control algorithm. Finally, we use example to illustrate the ARX model's precision and the efficiency of our data-driven model predictive control algorithm for general multirate sampled-data systems.