In the Decision Support System (DSS) model composition project, the single factor-driven method has a narrow range of application and low efficiency; the multi-factor-driven method lacks the description of the logical relationship between the algorithm and sub-model level and has low level of abstraction of the black box. Therefore, a composition method driven by the double factors of state and data was proposed. Based on the definition of workflow process model, the DSS submodel was generated as a workflow activity firstly, which was used as a portable and independent unit, containing the algorithm for solving the model inside and providing the external interface parameters. Secondly, according to the dependencies of activities performed, the corresponding Event-Condition-Action (ECA) rules were formulated to bind the composition relation between them dynamically. Finally, with the network activity diagram converted, combined work was completed by the aid of workflow platform. Experimental results show that the proposed method can solve the general problems of the model composition and has good generality and expansibility compared with other methods.