【摘要】 为了使近邻传播(AP)聚类在高维空间中获得更好的聚类效果,该文提出一种基于谱分析的近邻传播聚类方法(Affinity Propagation based on Spectrum analyze,AP-SA)。首先,通过采用谱分析技术将分布在高维非线性的数据点集映射到几乎线性的子空间上,映射过程实现高维数据降至低维。最后,通过AP聚类算法对映射在低维空间上的数据进行聚类,从而提高了AP算法在高维空间上的聚类性能。仿真实验结果表明,该方法相比于传统AP算法,在低维数据中无明显的优势,但随着实验的数据集的样本规模与维数的增加,在高维数据中的该方法降低了聚类时间的同时,也保证了较好的聚类效果。更多还原

【Abstract】 In view that affinity propagation(AP) clustering in high-dimensional space can get much better clustering results, this paper puts forward an affinity propagation clustering method based on spectrum analysis(Affinity Propagation-based on spectral analyze, AP-SA). First, by using the spectrum analysis technology, nonlinear data collection distributed in high-dimensional will be mapped to the almost linear subspace, and the high dimensional data is reduced to low dimensional in the mapping process. At last, through the AP clustering algorithm, the data mapped in low dimensional space is clustered, which improves the clustering performance of the AP algorithm in high dimensional space. The simulation results show that compared with the traditional AP algorithm, this method has no obvious advantage in the low dimensional data, but with the increase of the sample size and dimension of the experiment data, this method reduces the clustering time and ensures good clustering effect in the high dimensional data.更多还原