Abstract: New inferring methods of goal-oriented Credal networks are proposed such as based on genetic algorithms, Ant colony optimization algorithms and genetic-ant colony optimization algorithms are proposed for the large-scale Credal networks. Through designing, realizing and analysing these inferring methods, it is proved that these inferring methods can not only effectively obtain the best solution, but also can provide sensitive Credal sets of some variables influencing that specify status of goal variable for decision-makers in order to acquire the belief on the extreme decision.