Research On Attribute Reduction Algorithm Based On Dependency

Attribute reduction is the most in-depth problem of rough set theory.At present,there are deficiencies in the research of attribute reduction algorithms of rough set.For example,most reduction algorithms have high time complexity and can only deal with static data.Taking the classical rough set model as the starting point and the dependency between attributes as the heuristic information,this paper proposes an efficient algorithm for solving and reducing the data set.The specific contents are as follows:Firstly,in the traditional dependency based reduction algorithm,different attribute sets play different roles in the reduction process.If the dependency between the whole attribute set is traversed,it is bound to cause a waste of time and space.On the basis of extending the dependency method under equivalence relationship to the interdependence between attribute sets,this paper proposes a dependency ranking reduction algorithm,which only retains the necessary attribute sets for reduction.The comparative analysis of examples shows that the improved algorithm can effectively remove redundant attributes and obtain relatively better reduction results.Secondly,aiming at the situation that the classical reduction method based on attribute importance only considers the importance of a single attribute in the reduction process and ignores the correlation between attributes,which may lead to incomplete reduction results,the mutual information in information theory is introduced into the algorithm to measure the correlation of conditional attributes.On this basis,an optimal dependency reduction algorithm is proposed.The experimental results show that the new algorithm fully considers the correlation between attributes,and the reduction results are more complete.Finally,in view of the deficiency that most traditional reduction algorithms can only deal with static data,the relative dependency between attributes is used as heuristic information and integrated into the dynamic three way decision calculation model.Based on the increase of single sample in the dynamic data set,the three way regional update mechanism of sample increase is proved,and a three branch decision incremental attribute reduction algorithm based on relative dependency is proposed.The case analysis and UCI data experimental verification show that in the complex and dynamic big data environment,the improved algorithm has lower time complexity and higher reduction efficiency than the classical non incremental three way decision reduction algorithm.At the same time,the algorithm also has strong robustness and effectiveness.