Research On Attribute Reudction Of Distributed Set-valued Data Based On Rough Sets

With the diversification of data types,we may obtain nonstandardized data such as set-valued data due to the limitation of acquisition means or requirement of practical problems in practice.Accordingly,in order to better solve these problems,it becomes more and more important to study how to deal with nonstandardized data.Meanwhile,as a manifestation of big data,distributed data is commonly found in many practical problems.However,due to the limitation of high cost,large amount of data and security issues,it is usually impossible to realize centralized processing,so that the traditional methods are not effective anymore.Therefore,how to deal with the distributed data efficiently becomes a hot issue in current big data research.Attribute reduction is an important part of the research in data preprocessing,it can effectively remove redundant or unimportant attributes so as to accelerate the subsequent data processing.At present,researchers have put forward a large number of related researches and formed relatively complete results for attribute reduction of centralized data.For the distributed standardized data and distributed incomplete data,some researchers have proposed some corresponding theoretical methods to effectively remove redundant attributes on the premise of maintaining the claasification ability unchanged.However,attributed reduction for distributed set-valued data has not been investated yet.In this paper,attribute reduction for distributed set-valued decision information system was deeply discussed.The contributions can be summarized as follows:1.The attribute reduction of set-valued data under the disjunctive semantics is studied.Firstly,a new similarity relation based on probability is defined to measure the degree of similarity between two set-valued objects,which overcomes the limitations of existing methods.Then,a corresponding algorithm of attribute reduction based on positive domain is proposed for set-valued decision information system.The experimental results show that the proposed method can effectively simplify the set-valued decision information system,and compared with the existing methods,the classification accuracy is improved by using the proposed method in this paper.2.The attribute reduction for distributed set-valued data is discussed.Firstly,the definition of the rough set in distributed set-valued decision information system is given.Next,the attribute reducibility of the system is discussed on the premise of maintaining the positive domain unchanged,and an algorithm of attribute reduction for distributed set-valued information system is proposed.The experimental results indicate that the proposed method can effectively remove the redundant attributes in the distributed set-valued decision information system while keeping the classification ability of the system basically unchanged.