Research Of The Rough Set Model Based On Prior Probability Dominance Relation And Its Data Mining Methods In Incomplete System

The starting point of rough set theory is based on the existing knowledge,to classify the unknown information.Then it determine the degree of support of each partition to a certain concept.At the same time,the degree of support is represented by three approximate sets of positive,negative and boundary.Then,rough set is used to obtain decision rules by attribute reduction and simple algorithm.In this paper,a rough set model based on conditional prior probability dominance relation is proposed.It is established on the basis of the attribute value data statistics of incomplete partial order relation decision system.It not only takes into account the different conditions of the same attribute values,but also considers the correlation between different attributes,so that a variety of prior information can be fully utilized.Therefore,the classification accuracy and quality of classification can be improved effectively.This new model is proved to be effective and practical by theoretical analysis and practical example.Secondly,because the uncertainty measurement method based on knowledge granularity can not accurately and systematically reflect the uncertainty of the system,a new improved rough entropy based on boundary domain and knowledge granularity is proposed.The modified rough entropy not only takes into account the uncertainty caused by the inaccurate partition,but also take into account the changes brought by the boundary of uncertainty.So that the calculation of uncertainty measurement is more accurate.A new approach to the study of uncertainty measurement in the conditional prior probability dominance relation model is presented.Finally,reduction,distribution reduction and allocation reduction are introduced,and the relationship between the three and their properties are analyzed in detail.At the same time,the heuristic reduction algorithm based on improved rough entropy and the distribution reduction algorithm based on the target assignment matrix are proposed.Theoretical analysis shows that the latter is too cumbersome to reduce the search efficiency.In the process of reduction,the former can delete the unnecessary attributes in the system,so it can save the search time and improve the search efficiency.