Research On Three Branch Acceleration Method For Neighborhood Rough Set Attribute Reduction

With the development of information technology,data collected in various industries is growing explosively,and the uncertainty and fuzziness of the data are also growing.How to efficiently mine information and discover knowledge from complex data has become the focus in many fields,such as machine learning,data mining,knowledge discovery,etc.The attribute reduction method of the classical rough set model maintains the consistency between conditional attributes and decision attributes.It can reduce the redundant attributes in the data and compress the data scale.However,classical rough sets often cannot effectively reduce attributes in the face of highdimensional and continuous data in practical problems.Therefore,attribute reduction methods are gradually emerging based on neighborhood rough sets(NRS).Our work focuses on the attribute reduction method of neighborhood rough set as the essential starting point.(1)The existing attribute reduction methods based on neighborhood rough sets are too strict to deal with the information between neighborhood class and target set flexibly,which restricts the efficiency of attribute reduction.This paper proposes a three-way approximate reduction method based on attribute significance.Firstly,we introduce the cost loss function to analyze the similarity between neighborhood class and target decision class and divide all attributes into three subsets: positive,negative,and boundary sets;Secondly,we construct the neighborhood rough set model with the threeway approximation via shrinking the boundary setting;Finally,three-way approximate reduction sets are obtained by using the metric criterion of keeping the positive set unchanged to complete the attribute reduction.The experimental results verify the effectiveness and practicability of the reduction method.(2)The existing heuristic search methods in attribute reduction still have some problems: high complexity and time consumption.By introducing the sequential threeway decision model,we study attribute reduction methods' acceleration strategy and propose a sequential acceleration strategy based on the three-way approximate reduction.Firstly,based on the attribute significance,the attribute can be divided into three subsets: the positive set represents the accepted attribute,the negative set represents the rejected attribute,and the above operations are repeated for the delayed decision set until the reduction result meets the constraints;Secondly,according to the result of the simulation experiment,we can easily find that our method can effectively reduce the time consumption to ensure classification accuracy under the conditions of integration and globalization.Finally,the experiments conducted on the UCI benchmark show that the proposed model achieves favorable results.