Research On Attribute Reduction Based On Genetic Algorithm And Probability Rough Sets

Rough set,which has been applied to various fields currently as a new type of mathematical tool,is used to deal with some uncertain and incomplete information by making use of the indiscernible relation among the attributes in the set.As a result,the information in the system can be expressed more concisely and effectively.The hypothesis classification of the classical rough set model must be definite and it lacks flexibility in the practical application.Therefore,it has great limitations and cannot properly deal with the data with interference information.Based on the traditional rough set,many scholars put forward the concept of probability threshold and proposed the probabilistic rough set model.Attribute reduction is one of the major issues in the study of rough set theory.Objectively speaking,redundant information exists in various knowledge decision systems while attribute reduction is used to delete irrelevant attributes and simplify the attributes in the system.The so-called minimum reduction is a more concise and effective attribute decision rule.However,it is proved to be an NP problem whether the minimum attribute reduction can be worked out or not.Therefore,it is very necessary to find a proper attribute reduction algorithm in the daily life.Aimed at the uncertainty in attribute reduction process of probability rough set processing and combining with genetic algorithm,the thesis focuses on exploring the attribute reduction of probabilistic rough set,with two kinds of attribute reduction algorithms proposed.The main research results are as follows:1.Heuristic attribute reduction algorithm,based on probability rough set,necessarily calculates the updated value of probabilistic approximation accuracy after all attributes are added and deleted on the basis of attribute core,which has low efficiency because of the heavy calculation burden and large storage space.Based on this point,the reduction algorithm is proposed to improve the approximation accuracy and attribute of probability rough set.At the beginning of this method,the improved approximation precision and importance degree of each attribute are calculated.It is not necessary to calculate the approximation precision of all attributes after addition and deletion.According to the order of magnitude of the improved probability approximation accuracy,the minimum attribute reduction is obtained after the redundant attributes in the attribute reduction are removed based on the importance of attributes.Compared with the original algorithm,the efficiency and accuracy of simplification are enhanced to some extent.2.The fitness function of attribute reduction algorithm,based on the combination of traditional genetic algorithm and rough set,has some limitations and makes it necessary to check all subsets of an attribute set.In order to improve the accuracy and efficiency of attribute reduction,attribute reduction algorithm is proposed,which is connected with the genetic algorithm and based on genetic algorithm and probability rough set.The individual coding way of the initial population is limited by using the core attribute while the correction operator is joined to calculate the fitness.The method improves the original fitness function and regards conditional information content as the basis for judgment,correcting the direction of population evolution at any time and ensuring the diversity of the evolutionary direction and population.Finally,the reserved best individual is the minimum attribute reduction.The experimental results show that the reduction efficiency of the two attribute reduction algorithms is enhanced evidently.