Research On Fuzzy Rough Learning Algorithm Based On Information Fusion

Information fusion refers to the information processing process in which the data and information obtained from a single or multiple information sources are correlated and integrated,followed by a comprehensive and timely assessment of the situation,threat and their importance.Combining the idea of information fusion with the theory of fuzzy rough set,that is,using the fuzzy rough set to effectively integrate the multi-source information of things or objects,we can get a more comprehensive and integrated decision.Based on information fusion,this paper constructs the fuzzy rough attribute reduction algorithm and the classification model respectively.1.The fuzzy rough attribute reduction algorithm based on information fusion usually adopts the principle of maximum membership principle when making decisions by using the fuzzy rough set model.The classical fuzzy rough set decision model usually takes the mean value when calculating the decision lower approximation,which ignores the difference of the importance of each attribute to the information fusion.In order to solve this problem,firstly calculate the decision lower approximation for each attribute and then introduce the concept of fusion decision,and solve the decision lower approximation weight based on a single attribute through optimization theory.The weight is the fusion coefficient.The fuzzy rough approximation matrix is obtained by the decision lower approximation.Finally,the comprehensive evaluation based on the fusion attribute is made according to the maximum membership degree principle.At the same time,according to the calculated fusion coefficient,the fuzzy rough attribute reduction algorithm based on information fusion is constructed.2.The classification model under information fusion is based on the fuzzy and rough decision model of information fusion and takes the difference of decision into account for information fusion.The process of decision is basically the same as that in the previous chapter.However,due to the differences in decision classes,the related concepts and properties have changed.Firstly,the fuzzy rough decision approximation of decision class and its weight are redefined.The weight is the fusion coefficient.The fuzzy roughapproximation matrix is obtained by the decision lower approximation.Finally,the comprehensive evaluation based on the fusion attribute is made according to the maximum membership degree principle.In this chapter,two methods are used to solve fusion coefficient,namely Lagrange multiplier method and quadratic programming algorithm.