Research On Air Quality Evaluation Based On Decision Rough Set And Evidence Theory

With the rapid growth of economy,the environmental problems caused by it are more and more serious.In actual monitoring,air pollution data may be missing and redundant,which increases the difficulty of accurate analysis and prediction of air quality index.Faced with large-scale,high complexity,huge amount of information data,traditional data mining and research methods are facing great challenges.In essence,evidence theory and rough set are both effective mathematical tools for dealing with imprecise information,this paper proposes a decision rough set model in incomplete information system,and introduces evidence theory.Combining the two can better adapt to new needs and obtain more accurate decision-making information.When the recognition framework of evidence theory is incomplete,the concept of open recognition framework is proposed,and a new method of evidence synthesis is suggested,which solves the problem of basic probability assignment of unknown proposition and extends and improves Murphy average method to open recognition framework.The approach can effectively synthesize conflict data and further improve the evidence theory under incomplete recognition framework.The decision rough set model under incomplete information system is established by using maximum compatible class instead of equivalent class.For the problem of attribute reduction,this paper proposes two heuristic attribute reduction algorithms which keep the positive domain invariant for incomplete interval information system.In order to save resources,reduce the number of attributes in the reduction and improve the efficiency of the reduction.From a local perspective,this paper only performs attribute reduction for individual key decision-making classes,and proposes a local attribute reduction algorithm that preserves the positive domain.When data is fused,it is necessary to complete and unify data,reduce attributes and extract rules.After the preliminary processing of the information table data,the probability distribution function is redefined,so that the basic probability assignment function is determined,and the final decision result is obtained.Combined with the actual data of remote sensing and station observation,the single attribute fitting prediction,the trend prediction based on decision rough set theory and the trend prediction based on the combination of decision rough set and evidence theory are compared and analyzed.The rationality and validity of this method are verified by experimental results.