Introduction of the hybrid inference tool (HIT)
The construction of belief networks is a widely used methodology for high level fusion modeling. While some of the components of a belief network deal with ambiguous (probabilistic) data, others may deal with vague (possibilistic) data. Given the need to represent both probabilistic and possibilistic components in a single belief network, a framework and toolset for building Hybrid networks, utilizing Bayesian techniques to represent ambiguous data components and Fuzzy logic for vague components, is highly desired. To address this shortfall, we introduce the Hybrid Inference Tool (HIT) to aid in the construction, compilation and performing of inference on Hybrid networks. The design of HIT includes an Application Programming Interface (API) to support the incorporation of different transformation methods, engines and modeling packages for each component type. The applications utilized for the development, testing and validation of HIT are Norsys' Netica API (Bayesian networks) and the NRC CIIT's FuzzyJ API.
