相交网格下基于最优划分的多密度梯度网格聚类算法

为解决网格聚类算法中对参数过于敏感、无法自动识别不同密度梯度类以及不同梯度类间划分不够精确等问题,提出了相交网格下基于最优划分的多密度梯度网格聚类算法(OPMDG).该算法只需用户输入一个大致的密度阀值范围,网格边长自动计算并可自动调节适应,减少了算法对参数的敏感性;提出了二重划分技术,可挖掘不同密度梯度的类;对于处于不同类上的交界点,引入了电荷间吸引力的概念,能有效解决类间聚类精度不高等问题.实验结果表明该算法是有效的.

A Multi-density Gradient Grid Clustering Algorithm Based on the Optimal Division of the Intersecting Grid

In order to solve the grid clustering algorithm to parameter too sensitive,unable to automatically identify different density and different gradient differentiates precise enough to wait for a problem,the multi-density gradient grid clustering algorithm based on the optimal division of the intersecting grid(OPMDG) is put forward in this paper.The only user input to the algorithm of an approximate density threshold range,the grid side length automatically calculates and adjusts automatically to adapt to reduced sensitivity of the algorithm parameters;the two re-divided technology can mine the class of different density gradient;the introduction of the concept of charge attraction between the junction point in a different class,can effectively solve the clustering accuracy is not high in between classes.The experimental results show that the algorithm is effective.