Applying Bayesian Belief Networks in Rapid Response Situations

We have developed an enhanced Bayesian analysis tool called the integrated knowledge engine (IKE) for monitoring and surveillance. Our enhancements are suited for rapid response situations where decisions must be made based on uncertain and incomplete evidence from many diverse and heterogeneous sources. The enhancements extend the probabilistic results of the traditional Bayesian analysis by (1) better quantifying uncertainty arising from model parameter uncertainty and uncertain evidence, (2) optimizing the collection of evidence to reach conclusions more quickly, and (3) allowing the analyst to determine the influence of the remaining evidence that cannot be obtained in the time allowed. These extended features give the analyst and decision maker a better comprehension of the adequacy of the acquired evidence and hence the quality of the hurried decisions. We also describe two example systems where the above features are highlighted.