Intforbayes: uma ferramenta para sistemas especialistas bayesianos.

Abstract

This work describes IntForBayes, a "shell" for the development and consultation of special systems, whose main advantages are: intelligence in determining the order and number of questions to be asked (this intelligence corresponds to a "best-first search", with heuristic values calculated locally, reflecting the importance that the facts have to prove / disprove the chosen hypothesis); reasoning with chaining can be chosen between 100% progressive ("forward") to 100% regressive ("backward"); probabilistic inference based on Bayes' theorem, Dealing with uncertainty and incompleteness; pruning the research space; for the early investigation cycle of a hypothesis when a point of satisfaction is reached in the definition of its plausibility, and the possibility of "voIuntariamento" of information not requested by the system. The combined and intelligent use of all these features can be quite advantageous.