Abstract: Aiming at the various distribute clustering problems in diffusion model for all data points，a new clustering algorithm（CDD） based on the change of density is proposed.CDD searches the core point using a typical clustering algorithm（DBSCAN） based on the density，it calculates the direction，speed and acceleration of density diffused through analyzing the diffusion rule of data sample and its around the point’ density，then completes the sample points’ clustering.The experimental results show that compared with DBSCAN，CDD can cluster the diffusion model accurately，and has strong anti-noise-interference ability for the non-diffusion model which makes it easier to determine the merits of the parameters.

Key words: clustering, data mining, diffusion model

摘要： 针对各种扩散模式数据点分布的聚类问题，提出了一种基于密度变化的聚类算法（CDD）。CDD采用基于密度的典型聚类算法（DBSCAN）寻找核心点，通过分析数据样本及其周围点密度的扩散规律，计算密度扩散的方向、速度和加速度，对数据样本进行聚类。实验结果表明：与DBSCAN相比，能准确对扩散模式数据进行聚类，对非扩散模式数据具有抗噪声干扰能力强，参数较易确定的优点。

关键词: 聚类, 数据挖掘, 扩散模式