This thesis is concerned with the sensitivity analysis of rule-based expert systems. The types of sensitivity analysis include data driven, goal driven, and rule driven sensitivity analysis. The data driven sensitivity analysis investigates the impact of changing the known facts on the current conclusion. The goal driven sensitivity analysis shown which facts should be changed to move the current conclusion to another one. The rule driven sensitivity analysis is applicable when there exist multiple actions in OR relationship for a certain rule.
This thesis has developed a prototype, named SIS(Sensitivity Inference System). The SIS consists of knowledge base, fact base, inference engine, sensitivity analysis module, consistency check module, and user interface. The SIS is implemented on the microcomputer using the Turbo Pascal.