Increasing number of applications require predicting behavior of a complex system and making decisions to move the system towards desirable outcomes. Examples include finding the best course of action in emergency, deciding on business transactions within a supply chain, making a patient treatment plan for the best prognosis, and deciding on public policies guided by most positive outcomes.
In these applications, predictions and decisions are to be made in the presence of large amounts of dynamically collected data and learned uncertainty models. In many cases, it is also necessary to acquire additional data in order to reduce uncertainty and make better decisions. We call such a system, which supports a closed-loop data acquisition, learning, prediction and decision optimization, a decision-guidance application.
The focus of this research seminar is on studying models, languages, and algorithms toward building a decision-guidance management system (DGMS), which is a productivity tool for fast development of decision-guidance applications in a seamlessly integrated environment.
Significant advances have been made in the areas of operations research, mathematical and constraint programming, machine learning and data mining, and database systems. These advances can all contribute to a DGMS. However, there are no cohesive frameworks, algorithms and systems that unify the models and computational paradigms of all the components. A unified framework in the form of a DGMS, which is the focus of this class, is necessary for decision guidance in complex systems. In addition, the integration of multiple components from different areas in a unified DGMS brings new computational challenges and new optimization opportunities.
Class website: http://classweb.gmu.edu/brodsky/cs795infs797/