混合尺度聚类模型收敛性分析及仿真

Convergence analysis and simulation of mixed-scale clustering model

针对被分类对象属性多样性的客观实际,分析使用单一间隔尺度或名义尺度进行聚类的局限性,并提出一种混合尺度聚类模型。新模型以明氏距离及翻转距离为基础,引入混合尺度。利用下准则函数及逐步聚类算法可证新模型的收敛性。以某小区保安预警系统的采样数据为例,运用Matlab软件对新模型进行仿真验证,结果表明,新模型可应用于被分类对象具有多种特征属性的聚类问题。

In order to classify the objectives with complex attributes, a novel mixed-scale clustering model is proposed after analyzing the limitation of both interval scales and nominal scale for clustering.This model is based on Minkowski ’s distance and flip distance,and its convergence is proved with under criterion function and stepwise-clustering algorithm.Sample data collected form a security early-warning system of neighborhoods is used to verify the model with Matlab software.The results show that the model is valid to classify the objectives with complex attributes.