The Research On Multi?granulation Attribute Reduction Algorithm Based On Discernibility Matrix

With the development of technology and the arrival of big data era,the amount of data grows rapidly,which brings abundant data resources to human beings.However,there will be a lot of redundant information in these data,which makes the analysis and processing of data face many challenges.Therefore,research on how to extract effective information from data mining technology has become an important research direction.As an important tool of knowledge acquisition and data mining,rough set has been widely concerned all the time.Attribute reduction is one of the core contents of rough set theory.It can remove redundant and uncorrelated attributes,improve the classification accuracy while keeping the classification ability of information system unchanged.Multi-granularity is an important research direction in the field of granular computing.It can solve problems under different granularities and get more satisfactory and reasonable results.Multigranularity computing provides a new paradigm for solving complex problems.Therefore,this paper studies the attribute reduction algorithm from the perspective of multi-granularity.The main work is divided into the following two parts:(1)Aiming at the information system composed of symbolic data,the multi-granularity attribute reduction algorithm is studied based on discernibility matrix under the background of classical rough set.Firstly,the attributes are granulated based on the dependence of attributes on decision.Secondly,the importance of attribute granules and attributes in granule are defined based on discernibility matrix.The importance of attribute granules is used to measure the importance of the whole granules,and the importance of attributes in granule is used to measure the importance of attributes in granule.Finally,an attribute reduction algorithm is designed by using two important evaluation indexes,and the effectiveness of the algorithm is verified by experiments.(2)Aiming at the information system composed of numerical data,the multi-granularity attribute reduction algorithm is studied based on discernibility matrix under the background of fuzzy neighborhood rough set.Firstly,the correlation between attribute and attribute is calculated by Spearman formula,and then the attributes are clustered by spectral clustering to realize attribute granulation.Secondly,the corresponding discernibility matrices in the attribute sense and the positive region sense are defined respectively.Based on discernibility matrices,the importance evaluation indexes of attribute granules and attributes in granule are defined.Then,an attribute reduction algorithm is designed by using these two evaluation indexes.Finally,the experimental results show that the effectiveness of the multi-granularity attribute reduction algorithm.