Research On Accelerated Algorithm Of Attribute Reduction In Rough Sets And Its Neighborhood Model

With the rapid development of computer technology and explosive growth of data,the intelligent analysis of uncertain data has increasingly become a key scientific problem.Rough set is a mathematical tool to deal with uncertain,inaccurate and incomplete data.It only uses the information provided by the data set itself without prior knowledge,so it has a great development prospect in information processing.Attribute reduction is one of the research hotspots in rough set theory.Attribute reduction is one of the hot topics in rough set theory.Its essence is to exclude irrelevant or redundant attributes under the condition that the classification ability is basically unchanged.The purpose of attribute reduction is to greatly reduce the size of the data set,so as to use as few attributes as possible to carry as much data information as possible.However,the low efficiency of attribute reduction limits the development and practical application of rough set theory.Therefore,the research on fast attribute reduction algorithm has important theoretical value and practical significance.Based on the analysis and research of classic rough set model,neighborhood rough set model and common attribute reduction algorithms,this thesis puts forward two main research contents.Firstly,the properties of rough set theory are analyzed deeply,and the stability theorem of redundant attribute and the stability theorem of attribute local redundancy are proposed and proved.Based on these two theorems,a fast attribute reduction algorithm based on attribute local redundancy is designed.According to the relative active region of attribute and the stability theorem of redundant attribute,the algorithm not only gradually reduces the number of attributes involved in the calculation,but also gradually reduces the number of objects involved in the calculation during the iterative process of attribute reduction.Therefore,the algorithm can reduce the amount of calculation to obtain attribute reduction to a certain extent.The experimental results verify the effectiveness of the algorithm.Secondly,the classic rough set model can only deal with symbolic data,but cannot deal with continuous data directly.To solve this problem,experts and scholars have extended the classic rough set,and the neighborhood rough set model is one of the important extensions.The neighborhood rough set needs to repeatedly calculate the neighborhood of each object in the process of attribute reduction,which has high computational cost and long running time.In this thesis,combined with ball clustering,the entire universe is clustered to generate a series of ball clusters.In the process of attribute reduction,the neighborhood of each object is not repeatedly calculated,but the corresponding neighborhood of the generated ball cluster's centroid is calculated.Therefore,a fast attribute reduction algorithm of neighborhood rough set based on ball cluster is proposed.Experimental results show that the algorithm consumes relatively less time.