差转计算算法在连续型因素上的改进与应用OA北大核心CHSSCDCSTPCD

In order to solve the problem of low degree of reliability of inference knowledge mined by the Set Subtraction and Rotation calculation(S&R),poor generalization of knowledge,and certain judgmental risks in the generalization process based on continuous-type data,in the context of probability theory and mathematical statistics and factor space theory,this paper combines with the S&R calculation principle to propose a new discrete method for continuous type data and construct the cumulative deter-minant,relative contribution for measuring the confidence level of inferred knowledge.In order to validate the effectiveness of the discretization method,it is combined with the S&R calculation and applied to the auxiliary diagnosis of malignant tumors,and the decision tree is used as the comparison algorithm.The empirical results show that the fusion of the proposed discretization method and the S&R calculation improves the generalization effect of the computation algorithm effectively,that the overall performance of the fused algorithm decision-making is comparable to that of the decision tree,but the knowledge representation is simpler than decision tree,and that the two constructed indicators of cumulative determination and relative contribution are able to effectively measure the contribution and trustworthiness of inferred knowledge.