Acquiring Probabilistic Knowledge for Information Fusion

Acquiring probabilistic knowledge for the development of fusion models can prove to be a difficult task. In many applications, the lack of availability of historical data forces the modelers to rely extensively on Expert Judgment techniques which traditionally requires a large amount of manpower and is exceedingly time-consuming. The DAC team proposes a novel approach to the knowledge elicitation process which will address these limitations through the development of an automated prototype system for the collection and analysis of probabilistic knowledge from Subject Matter Experts. Using traditional survey and estimation techniques in combination with non-parametric statistical tests, we aim to provide modelers with a computerized system for the development of Bayesian Inference Models to perform high-level information fusion. This method will build on DAC’s extensive expertise in Bayesian network design and integration with automated systems. The results of this effort will be an intuitive, robust prototype for knowledge elicitation which will pave the way for an online deployable tool to be implemented under future research.