基于信息源选择的多源信息系统邻域互信息熵属性约简

Neighborhood Mutual Information Entropy Attribute Reduction of Multi-source Information System Based on Information Source Selection

属性约简是粗糙集理论的重要研究内容。针对多源信息系统中信息源之间存在相互依赖的问题,提出一种基于信息源选择的多源信息系统邻域互信息熵属性约简算法。首先提出多源信息系统下信息源的差异度度量,通过该度量可以评估信息源在整个多源信息系统的独立程度;接着利用差异度对多源信息系统的信息源进行选择,使得选择出的信息源具有较高的独立性;最后基于邻域互信息熵模型,对选择出的新多源信息系统进行属性约简,提出相应的属性约简算法。实验分析表明,所提出的多源信息系统属性约简算法具有较高的有效性和优越性。

Attribute reduction is an important research content of rough set theory.Aiming at the problem of interdependence between information sources in multi-source information system,a neighborhood mutual information entropy attribute reduction algorithm of multi-source information system based on information source selection is proposed.Firstly,the difference measure of information sources is presented in multisource information system,through which we can evaluate the independence of information sources in the whole multi-source information system.Then the difference degree is used to select the information sources of the multi-source information system,so that the selected information sources have high independence.Fi⁃nally,based on the neighborhood mutual information entropy model,the attribute reduction of the selected new multi-source information system is carried out,and the corresponding attribute reduction algorithm is al⁃so proposed.Experimental analysis shows that the proposed attribute reduction algorithm for multi-source in⁃formation system has high effectiveness and superiority.