一种自动确定密度峰值的聚类算法改进

相比较于其它聚类算法，密度峰值聚类算法可将任意形状的数据与较少的参数和高效的聚类速度结合起来。针对当某个类中出现多个密度峰值时，聚类结果缺乏准确性的问题，提出一种改进的密度峰值聚类结果有效性造成的影响，算法通过比较类簇之间的密度属性，实现动态的子簇合并，减少主观因素对算法结果的影响。通过实验与已有密度聚类算法对比，改进算法不仅很好地避免了原算法人为确定参数给实验结果造成的影响，而且具有更好的聚类性能。

An Improved Clustering Algorithm for Automatically Determining Density Peaks

Compared with other clustering algorithms， the density peak clustering algorithm can combine arbitrary shape data with less parameters and efficient clustering speed. In order to solve the problem of clustering results when there are multiple density peaks in a certain class， the improved density peak clustering algorithm is proposed. The algorithm chooses the cluster center by considering the change of the angle between three points （the current data point， the previous data point of the current point and the last data point of the current point）， based on the change of the value of the decision point. At the same time， in order to reduce the influence of human factors on the validity of clustering results， the algorithm achieves dynamic sub-cluster merging by comparing the density attributes between clusters， reducing the influence of subjective factors on the algorithm results. Comparing with the existing density clustering algorithm， the improved algorithm not only solves the effect of the original algorithm on the experimental results， but also has better clustering performance.