Research On Rough Set Attribute Reduction Algorithm And Classifie

Rough set is a mathematical idea used to classify uncertain knowledge.It has been comprehensively developed today.Its theoretical results have been used in industrial production,data processing,medical education and many other aspects,it has a good reputation and application performance.Under the research of relevant scholars,the rough set model extends many different branches such as fuzzy model,decision model,variable precision model,etc.,and expands the application of rough set theory.Neighborhood rough set is an improved algorithm of rough set model.The concept of granulation and neighborhood space is introduced through theory,so that the model can be applied to continuous data,which solves the limitation that the original model can only deal with discrete variables.Attribute reduction is the core idea of rough set theory.It is used to filter important information from knowledge systems and remove useless information.It can obtain the key information we want without reducing the classification accuracy and maintaining the validity of the data system.This paper mainly researches and improves the attribute reduction algorithm of traditional rough sets and neighborhood rough sets,and validates the effectiveness of the algorithm through experiments,and combines the improved algorithm with the decision tree classification algorithm to optimize the classification performance and efficiency.The research work of this paper is as follows:(1)The commonly used classical rough set algorithms are introduced.Aiming at the problems that traditional rough set algorithms cannot obtain ideal results when analyzing large-scale data,an improved attribute reduction algorithm based on mutual information is proposed.In order to improve the accuracy of the algorithm,the calculation of conditional entropy is introduced;the importance of attributes is considered by conditional probability,which reduces the complexity of the algorithm;the data compatibility is prioritized,and the number of sample calculations is optimized.Using multiple data sets to conduct comparative experiments,it is proved that the algorithm in this paper can obtain better reduction rules and obtain ideal reduction results.(2)The algorithm idea and development process of neighborhood rough set are introduced,and a new attribute reduction algorithm is proposed combining with the idea of block set neighborhood rough set.Through the probabilistic sample attribute selection,the sample size required for calculation is reduced;through the reserved attribute reduction optimization algorithm,the calculation process is stored and the time consumption of the algorithm is reduced.The feasibility of the algorithm is verified by experiments.(3)The commonly used classifier algorithms are introduced,the core idea of the CART algorithm is expounded,the neighborhood reduction algorithm proposed in this paper is combined with the CART algorithm,the data preprocessing is performed,and an improved algorithm is proposed.After relevant experiments,it is proved that the improved algorithm can obtain better algorithm performance without affecting the classification accuracy.