Key features for model building decision support systems

We present a two-phased approach to automating the model building process in decision support systems (DSSs). The role of problem structuring in the model building process is stressed and is related to basic model building blocks (model types), which are in turn related to problem solving techniques developed in the artificial intelligence field. We propose the combined usage of classification and constructive problem solving methods. The classification method provides a high-level structure to the model building process since it facilitates the identification of the relevant model types. The constructive method then adapts and integrates the identified model types. Thus, the proposed approach ultimately constructs a model corresponding to a given problem but in a deferred fashion. The advantages of this two-phased model building process are pointed out and related to the findings of cognitive studies. The proposed approach is illustrated using the linear programming (LP) modeling domain with examples from the production planning application domain.