Study Of Inference Rules And Deduction Process On Decision Implications

Knowledge reasoning is a key technology of knowledge engineering,playing an important role in the field of artificial intelligence and showing great potential in many practical applications.Decision implication is a form of knowledge representation,with clear knowledge structure and underlying logic,and being easy to derive new knowledge through deduction.At the same time,decision implication has complete semantic foundation and can also guarantee the correctness of deduction.Therefore,in recent years,many scholars have focused their attention on the deduction of decision implication.At present,there is little research on inference rules and how to efficiently apply inference rules to derive new decision implications.Therefore,in order to make deduction more efficiently,it is necessary to study whether there exist more concise inference rules and how to use them efficiently to derive new decision implications.This is the reason why this paper starts with designing more concise and effective inference rules,and exploring the optimal deduction strategy.The contributions of the paper can be summarized as follows:(1)Propose a more concise inference rule-Consequence Combina-tion.This paper simplified Combination to Consequence Combination,that is,only the decision implications of the same condition attribute should be merged.Furthermore,we analyzed the relationship between Augmentation,Combination and Consequence Combination,and proved that Combination is redundant to Augmentation and Consequence Combination.Therefore,a new set of inference rule,including Augmentation and Consequence Combination,was proposed and proven to be complete and non-redundant.In addition,the properties of the new inference rule were also studied,especially the interchangeability between the inference rules.(2)Propose three inference methods,and verify the effectiveness of the inference methods by experiments.We studied the number of times of applying inference rules,and proved that Combination and Consequence Combination should be applied at most[log2m]times.Based on this result,we proposed three inference methods and verified their effectiveness through experiments.The experimental results show that applying Augmentation once,and then applying Consequence Combination at most[log2m]times is the most effective inference method at present.