结合灰狼优化算法和动态邻域的三支密度峰值聚类算法

A Three Way Density Peak Clustering Algorithm Combining Grey Wolf Optimization Algorithm and Dynamic Neighborhood

针对密度峰值聚类算法聚类效果受截断距离dc的取值影响较大以及传统二支聚类处理不确定数据强制划分带来的决策错误,提出结合灰狼优化算法和动态邻域的三支密度峰值聚类算法。首先,为解决截断距离dc的选取难问题,将聚类内部指标Silhouette指标作为目标函数,利用灰狼优化算法(GWO)的全局寻优能力求解最优的截断距离dc;为了使不确定数据的划分更加合理,结合动态邻域的思想,利用K近邻算法将二支聚类结果转化为三支聚类结果。通过在人工数据集以及UCI真实数据集的实验验证,该算法的聚类精度和总体性能优于其他5种对比算法。

The clustering effect of density peak clustering algorithm is affected by the truncation distance.The value of c has a significant impact and the decision errors caused by the forced partitioning of uncertain data in traditional two-way clustering processing.Therefore,a threeway density peak clustering algorithm combining the Grey Wolf optimization algorithm and dynamic neighborhood is proposed.Firstly,to address the truncation distance.The problem of difficult selection of c is to use the Silhouette index as the objective function within the cluster,and use the global optimization ability of the Grey Wolf Optimization Algorithm(GWO)to solve the optimal truncation distance;In order to make the division of uncertain data more reasonable,combined with the idea of dynamic neighborhood,the K-nearest neighbor algorithm is used to convert the two-way clustering results into three-way clustering results.Through experimental verification on artificial datasets and UCI real datasets,the clustering accuracy and overall performance of this algorithm are superior to the other five comparative algorithms.