Uncertainty Management for Dynamic Decision Making

Major Goals There are two goals in this project 1. Measure uncertainty in beliefs considering multiple root causes of uncertainty We extend an existing belief model, called Subjective Logic SL, to measure uncertainty in beliefs based on the root causes of uncertainty, including lack of information or knowledge, vagueness, and ambiguity. Although SL is developed to deal with the dimension of uncertainty explicitly unlike the existing belief models, SL considers uncertainty derived from lack of information and vagueness but does not deal with ambiguity derived from conflicting evidence. 2. Reduce uncertainty in data with high scalability We develop scalable uncertainty reduction algorithms by identifying the minimum set of data features to maximize decision effectiveness. A decision-maker often loses high utility by delaying a decision due to numerous alternative decisions with an equal utility or a high volume of uncertain evidence. Our goal is to significantly reduce uncertainty using a minimum set of features while maximizing classification accuracy.