采用聚类特征的基本概率分配生成方法及应用

A Method to Generate Basic Belief Assignment Based on Clustering Analysis and Its Application

针对在识别框架不确定时基本概率分配(BBA)生成困难的问题,提出一种基于聚类特征的基本概率分配生成方法,以减弱对样本长度的依赖性,并分析2种情况下的BBA生成。在框架未知时,通过聚类分析获得各个类别的聚类特征,建立样本属性的聚类特征区间模型;在框架已知时,获取聚类特征,建立样本属性的聚类特征区间模型;然后用各个区间模型之间的距离表示样本属性之间的差异,在此基础上建立了一种相似度的度量方法;最后对相似度进行归一化得到BBA。采用Iris数据集和Wine数据集的实验结果表明:所提方法对样本长度敏感程度低,对Wine数据集的一个类的分类结果达到100%。将该方法应用于某煤化工企业压缩机组子系统状态监测信息数据集,实现了监测信息状态的识别。

A method to generate BBA （basic belief assignment） based on cluster analysis is proposed to focus the problem that the mass function is hard to determine when the frame is unknown. The method tackles the situation whether the frame of discernment is known or not. A clustering analysis method is applied to extract cluster features and models of cluster features are constructed with the samples. Then the distances between different cluster feature models are calculated to represent differences between sample attributes and then the similarities of them are obtained. Finally, the values of similarities are normalized to get the BBA. The analysis results of classifying the Iris dataset and Wine dataset show that the proposed method is less dependent on the length of samples and the classification accuracy in Wine dataset is 100%. Monitoring information series by applying the method to a compressor unit system proves the effectiveness of the method, and the condition of monitoring information can be clearly recognized.