Research On Optimization Method Of Hybrid Attribute Reduction Based On Neighborhood

Attribute reduction belongs to the research field of data mining.It is a method to delete redundant attributes without reducing the ability of data set classification.In the application research of attribute reduction,in order to solve the problem that the classical rough set can not effectively deal with the numerical attribute data,people propose the neighborhood rough set,which can directly deal with the numerical and symbolic attribute data at the same time.However,the research on neighborhood rough sets generally focuses on numerical attribute data,and there are relatively few studies on the mixed attribute data that exist in real life.Therefore,for the hybrid attribute information system,this paper conducts an in-depth study on the attribute reduction of the neighborhood rough set,and mainly does the following optimization work:(1)Aiming at the problem of insufficient and inaccurate attribute reduction due to ignoring decision-making boundary objects in traditional neighborhood rough set attribute reduction,a hybrid attribute reduction algorithm HARNMCB(Hybrid Attribute Reduction based on Neighborhood Maximum Compatible Block).The algorithm selects the largest neighborhood maximum compatible block in the ?neighborhood compatible block of the universe object to construct the neighborhood rough set,and then performs attribute reduction.The experimental results show that compared with the traditional neighborhood rough set attribute reduction algorithm,HARNMCB can obtain a more effective attribute reduction subset.(2)Aiming at the problem that the accuracy of the neighborhood approximation is not accurate enough to describe the uncertainty of the neighborhood information system,the two uncertainty measures with complementary relations are combined,and an uncertainty measure based on the weighted integration idea is studied to evaluate the uncertainty.Attribute importance,and from this,a new uncertainty measure WIUM is proposed,and a hybrid attribute reduction algorithm HARWIUM(Hybrid Attribute Reduction based on Weighted Integrated Uncertainty Measure)is designed.The weighted integrated uncertainty measurement WIUM also considers the uncertainty of the data set and related knowledge,and also considers the impact of the size of the decision-making class on the uncertainty measurement results of system classification and decision-making.UCI experimental results show that compared to the uncertainty measurement that only considers a single factor,the weighted integrated uncertainty measurement method HARWIUM is significantly better in terms of the effect of hybrid attribute reduction for neighborhood rough sets.