| Automatic reasoning is an important field in Artificial Intelligence. In conventional common sense reasoning, new knowledge is inferenced from existing knowledge according to classical logic. However, in real environment, knowledge is often uncertain, inconsistent and incomplete, often referred to as the belief. According to the principle of classical logic, the conclusion is not significant if the condition of reasoning is inconsistent. Therefore, in order to make the reasoning useful, people put forward a variety of ways to solve the reasoning of inconsistent knowledge, in which the belief revision is a common one. The main idea of belief revision is to sort the belief set according to the degree of preference, to delete the part which causes inconsistency and is the lowest preferred.Compared to the belief revision, the non-revision reasoning does not remove the elements of the belief set. Instead, the new inference method is proposed to obtain a unique, consistent, closed set of conclusions. The belief revision method preserves the original belief set and reduces the loss of information. In this paper, we propose a new method of non-revision reasoning with clause in the predicate logic, proving the uniqueness, consistency and conductively closure of the result set. At the same time, this paper discusses the Epistemic Process of the non-revision reasoning in predicate logic, and proves that the process is convergent.At last, this paper proposes a classification system based on prior knowledge and rough set rules extraction method, which combines rough set rule extraction method and non-revision reasoning method. Compared with the classification system based on rough set rules extraction, the classification result of this system is more reasonable. At the same time, the system can extract knowledge from the different forms of decision system and collect them to a knowledge base, which improves the efficiency of information exchange, and protects the privacy of information. |
PDF Full Text Request