This paper proposes a data-based characteristic model of complex networks that is suitable for control design and analysis. A compression method called characteristic modelling is used to reduce the complexity of dynamic networks and the model with topology-structure preservation is proposed to describe the networks in a concise and explicit way. Based on the controllability and observability of complex networks, the minimal dimensional characteristic model is constructed for control design, which captures the major information of the networks. This modelling method also has the advantage of modelling changeable networks and finding the optimal driver nodes to control the networks. The application to complex networks show the effectiveness and easy implement of the characteristic modelling method.