邻域粗糙集在属性约简及权重计算中的应用

Application of attributes reduction and weights calculation through neighborhood rough set

为了减少航天器特征值属性的冗余性并提高其权重的准确性,提出了一种基于邻域粗糙集的属性约简及权重计算方法。通过对不同重要度下限分类精度的对比分析,给出了确定邻域半径的新规则。在信息观权值最优计算公式的基础上,提出了一种基于信息熵的特征值权重计算方法;给出了代数观和信息观最优组合权值确定方法,解决了代数观和信息观方法的权衡问题。将其应用于某卫星姿控系统特征值分析中,与其他方法的比较表明该方法能有效减少特征值的数目,提高特征值权重的准确性。

In order to reduce the redundancy of spacecraft attribute value and improve the accuracy of weights calculation, a new method based on rough sets is put forward. By comparing classification accuracy in different lower bounds of importance degree, the new rules for determining the neighborhood radius are given. Based on optimal formula to calculate weights in the view of information, a method to calculate weights of attributes based on information entropy is proposed, and a method for determining weights based on optimal combination of the algebra view and information view is put forward. The balance problem between algebra view and information view is solved. The method is applied to analyze characteristic values for a satellite attitude control system. A comparison with other methods shows that the method can effectively reduce the number of characteristic values and improve the accuracy of feature weights.