论文部分内容阅读
针对self-tuning谱聚类算法采用自适应高斯核计算相似度的方法对一些复杂结构的数据集无法正确聚类的问题,提出一种基于密度系数和共享近邻的谱聚类算法.首先计算每个样本点的密度系数,由密度系数阈值计算样本点的权值和加权的自适应核参数;然后根据密度系数阈值优化互为K近邻图并计算样本点之间共享近邻点的个数;最后根据核参数和共享近邻点的个数计算所有样本点之间的相似度并进行聚类.分别在人工数据集和真实数据集上进行实验,结果表明本文方法在处理一些复杂结构的数据集时可以得到更优越的聚类效果.“,”The self-tuning spectral clustering algorithm uses the adaptive Gauss kernel to calculate the similarity cannot get correct results on complex datasets,a spectral clustering based on density coefficient and shared nearest neighbors is proposed in this paper.Firstly,the density coefficient of the points are calculated and the adaptive kernel parameters are calculated based on the weight.Then,the mutual K nearest neighbor graph is optimized based on the threshold value and the number of shared nearest neighbors are calculated.Finally,the similarity is calculated based on the number of shared nearest neighbors and kernel parameters and clustering.Experiments on artificial and real-world datasets show that the proposed algorithm can obtain a better clustering result in dealing with the complex datasets.