Research On The Influence On Credibility Of Reasoning Results Caused By Relevant Evidence

In the design and development process of some expert systems, we usually find that the proof is relevant, such as the two rules that if{a, b} then D and if{b, c} then D,{a, b} and{b, c} are obviously related. In this case traditional inexact reasoning method is not useful for the fusion of results credibility any more, because they can only deal with the situation that the proof is independent of each other. The aim of this paper is finding a way to make the traditional inexact reasoning method apply to the situation that the proof is relevant, so that the reasoning results become more accurate.This paper introduces some related knowledge of inexact reasoning and expert system at first. Then it focuses on its theoretical research results. By analyzing relativity of proof, this paper puts forward a definition for the relativity of proof which can quantify the relativity. It improves credibility theory by introducing the relativity factor into the formula for the fusion of reasoning result, so that the formula becomes able to deal with the situation that the proof is relevant. At the same time the validity of this improved credibility theory is verified by a certification in theory. According to the improved credibility theory presented in this paper, this paper designs and realizes an archetype of expert system that can deal with the situation that the proof is relevant in the end. The test result shows this system can effectively resolve the problem that how to combine two results when the proof is relevant. Its reasoning results are more accurate than those traditional expert systems’which can’t deal with the situation that the proof is relevant.