中文摘要：

对近邻传播聚类算法进行了详细的研究。提出用波动率来衡量数据震荡的剧烈程度，分析了收敛迭代次数和最大迭代次数两个参数的合适设置；重点研究了倾向度和阻尼因子两个参数与聚类数、波动率的关系，研究发现通过增大倾向度和阻尼因子都能减少波动率。

英文摘要：

The affinity propagation clustering was studied. The volatility was proposed to measure the nu- merical oscillations＇ intensity,and the convergence of iterations and the maximum number of iterations of the right to set two parameters were analyzed. The research focus on the relationship of preference and damping factor with clustering and volatility, and the results indicate that it can reduce volatility by increas- ing the damping factor and preference.