Research On Classification Method Of Inconsistent Information Systems Based On Dominance-based Rough Set

With the rapid development of information technology,especially the rapid development and extensive use of database systems and network technologies,the amount of information stored in the information system is getting larger and more complex,increasing the difficulty of using information resources.In real life,the decision tables obtained are inconsistent,inaccurate and incomplete due to various operations and processing methods on the collected data.How to draw useful information from these uncertain information systems has become an important issue in the research of intelligent information systems.Rough set theory is a new tool for dealing with uncertain information.At present,the treatment of inconsistent information is mainly based on the variable precision and dominance rough sets model,but the selection of a scientific and reasonable variable precision threshold is the key.Most of the access to the variable precision thresholds is given through many experiments or received by field experts,but there is no priori available for some unknown fields such as spaceflight and deep sea that need to be explored in new ways.Therefore,when the data is uncertain and there is no priori knowledge available for reference,how to get rid of the dependence on prior knowledge in the learning process and make data-driven autonomous learning become a difficult problem in machine learning.In this paper,for the inconsistent information system,we do the following two parts from the rough set model of variable precision advantage relation and the threshold selection of extended rough set model.(1)Aiming at the inconsistent information system,a metric to measure inconsistent information is given.Based on this,a data-driven autonomous learning algorithm is improved.The core idea of this algorithm is to avoid the dependence on prior knowledge,and only by using the information of the data itself,the threshold of variable precision can be obtained,so that the rules can be effectively extracted.(2)When there are many attributes in information system,the definition of the traditional dominant rough set is too strict.Therefore,for the extended variable-precision rough set model,the autonomous learning algorithm is applied to it,which makes the processing of preference information more effective,and increases the extraction of useful information to improve the classification performance of the model.(3)For the above two parts,two sets of data from UCI database were selected to conduct two experiments to verify the feasibility and validity of autonomous learning algorithm,as well as to demonstrate the advantages of the extended relational rough set compared to the the superiority of traditional dominant relational rough set.