Target Optimization

Overview

Target Optimization searches for the sets or sequences of evidence (Hard Evidence and/or Numerical Evidence) that best optimize the probability of the target state or the mean value of the target node. The best combinations are scored and saved in the Evidence Scenario file.

The reports describing the optimization solutions provide several metrics, including the Posterior Probability $P(s \mid E)$, where s is the state to be optimized; the Marginal Likelihood $P(E)$, where E is the current set of evidence; the Likelihood $P(E \mid s)$; the Bayes Factor $BF(s, E) = P(s \mid E)/P(s)$; and the Generalized Bayes Factor $GBF(s, E)$. When the optimization targets the mean value, the report instead reports a Confidence Interval, whose confidence level can be modified via Preferences.