Research On Attribute Importance Measure Theory And Method Based On Data Coordination

With the rapid development of modern science technology and information technology,the data information in database becomes huge.Therefore,the importance and value of data have been paid more and more attention and become a hot research content in academia and industry.As a mathematical tool to deal with inconsistency,incompleteness and other uncertain information,rough set theory can effectively discover knowledge and potential rules hidden in data.For roughness measurement problem: On the basis of analyzing the characteristics and shortcomings of current variable precision roughness measurement method,the concept of synthetic effect function is proposed,and a roughness measurement model based on the effect precision is established(EPRD).The characteristics of EPRD are discussed through theoretical proof and specific example.Then,an attribute reduction method based on EPRD is proposed(EPRD-RM),and the similarities and differences between EPRD-RM and other attribute reduction methods are further analyzed with several commonly used UCI data sets.In addition,according to the decision information system,the concept of core data set is given,and the decision roughness measurement based on the data effect is established(DE-DRD).The performance of DE-DRD is analyzed through theoretical proof and specific example.Then,an attribute reduction method based on DE-DRD is proposed(DE-DRD-RM),and the necessity and feasibility of DE-DRD-RM are further analyzed with several commonly used UCI data sets.For attribute importance measurement problem: On the basis of data system,taking the knowledge hidden in data system as the carrier,and based on the inclusion relationship between sets,the concept of system decision coordination degree is presented,and a composite measurement of attribute importance based on core data is established(BCD-AIM).Then,the basic properties of BCD-AICM are discussed,and the features of BCD-AICM are analyzed with specific example.Finally,the similarities and differences between BCD-AICM and existing attribute importance measures are discussed with several commonly used UCI data sets.Therefore,the models in this paper have good explanatory and structural features,which can enable managers to make reasonable decisions in an uncertain environment,enrich the existing relevant theories to a certain extent,and have broad application prospects in the fields of fuzzy decision-making,knowledge acquisition,resource management,artificial intelligence and so on.