为了解决网格聚类算法中的输入参数和聚类结果不精确问题,提出了基于局部密度的动态生成网格聚类算法(DGLD)。该算法使用动态生成网格技术能大幅度地减少数据空间中生成的网格单元的数量,并简化邻居的搜索过程;采用局部密度思想解决数据空间相邻部分对网格密度的影响,提高了聚类精度。该算法不需要用户输入参数,能识别任意形状的聚类并有效地去除噪声点。实验结果表明该算法是有效的。
To resolve the problems that the grid-based clustering algorithms need the input parameters and the cluster results are imprecision, the dynamic-creating grids clustering algorithm based on local density (DGLD) is put forward. The algorithm uses dynamic-creating grids technique to reduce greatly the number of grid partitioned in data space, and simplify the neighbor search process. The algorithm adopts local-density idea to deal with the influence of the neighbor part to the current grid in data space. The algorithm doesn't need user to input parameters, and discover cluster of various shapes and eliminate noises effectively. The effectivity of the algorithm is shown in this experiments.