Research On Attribute Reduction Approach Based On Attribute Processing Mechanism

With the continuous development of information technology in the era of big data,humanity is facing the challenge of extracting high value information from the massive data generated daily in various industries,which exhibits characteristics such as large-scale,complex-structure,and low-value density.The rough set method has been proven to be able to process massive data and has been widely used in data mining and other fields.With further research,many scholars have proposed extended models such as neighborhood rough set,fuzzy rough set,and supervised neighborhood rough set to address the limitations of the classic rough set model in directly handling continuous data.Regardless of the model used,attribute reduction has always been an important research direction in the field of rough set,with the aim of eliminating redundant attributes in the data,and output a minimum subset of attributes that meet the pre-set conditions,thus reducing the dimension of the data.However,the traditional heuristic method for computing reduct is time-consuming,which poses difficulties for subsequent learning when dealing with data.Therefore,this paper will focus on how to improve the efficiency of solving attribute reduct from the perspective of attribute processing and design two methods for quickly solving attribute reduct.Specifically,the main research content and methods of this paper include the following two points:1.A method for accelerating attribute reduction based on attribute dissimilarity is proposed.In fuzzy rough set,the Gaussian kernel function is introduced to measure the similarity between samples and construct the corresponding fuzzy relation.However,in the process of computing reduct based on fuzzy rough set using heuristic method,each candidate attribute needs to be iteratively traversed,which will lead to more time consumption.To address this issue,this paper utilizes the dissimilarity between attributes and employs the fuzzy granular structure distance to partition the conditional attribute set into different combinations.During the process of searching for reduct,multiple attributes that satisfy the conditions are selected and added to reduct set based on attribute combinations,this process reduce the number of attribute searches and improve the efficiency of computing reduct.Finally,experimental results on 12 UCI datasets demonstrate that the proposed method effectively reduces the time consumption of computing reduct while ensuring that the obtained reduct has comparable classification performance.2.A method for accelerating attribute reduction based on attribute pre-sort is proposed.Supervised neighborhood differs from traditional neighborhood by considering the similarity of samples based on their class labels.It uses the intra-class and inter-class radii to determine the similarity of samples with the same and different class labels,respectively,reducing the uncertainty in the neighborhood.However,heuristic method is used to compute reduct procedure based on supervised neighborhood rough set when the number of attributes is large,the time efficiency will be reduced.To address this issue,this paper fully explores the inherent information between condition attributes and decision attributes before reduction.Specifically,the importance of each attribute is evaluated based on the ratio of dispersion between samples of different classes and aggregation between samples of the same class under the attribute,and the attributes are then ranked according to their importance,and the reduction process only selects the attributes that meet the condition in order.Finally,experimental results on 10 UCI datasets demonstrate that the proposed method not only improves the efficiency of computing reduct but also provides satisfactory classification performance.