Abstract: This study proposed a new algorithm based on BAYES rule, after deeply researching the reasoning methods which related with uncertainty problems of expert system. This algorithm introduced the thought of credibility, by establishing inference network model, started from the known observed evidences to forward reasoning and eventually got the probability of the impact of evidence on object conclusion. And it also greatly reduces the probability of which uncertainty faction caused object conclusion distortion. The reasoning result showed it works well and meet expectation.