Dynamic Service Composition for Supply-Chain Management Decision Support

Complex decision-support systems can be very effective for data analysis and for solving combinatorial problems. But to effectively support real-world business decisions, systems must do more than track data and solve problems. They must also be transparent and user-adaptable. We will describe a service-oriented approach that allows rich analysis and modeling tools to be composed with easy-to-understand dataflow components. We have used this architecture to build an autonomous trading agent for the Trading Agent Competition for Supply-Chain Management (TAC SCM). But since full autonomy is often not practical or desirable for real-world business situations, we also show how this approach can be used to construct "dashboards" that allow a user to view and interact with data flows, to visualize the dataflow network, and to modify the network in real time. We make the case that formal semantic descriptions of individual dataflows and services must be significantly richer than the typical declarations of data types that are commonly used by data analysts and software developers. We have implemented this approach to service description and composition using the Semantic Web Resource Description Framework coupled with readily-available inference tools.