结合蝙蝠算法改进的密度峰值聚类算法

Improved density peaks clustering algorithm combining bat algorithm

密度峰值聚类算法(Denisity peaks clustering,DPC)具有聚类速度快、实现简单、参数较少等优点,但该算法的截断距离参数需要人工干预,并且参数的选取对于该算法的结果影响较大。为了解决这一缺陷,该文提出了结合蝙蝠算法改进的密度峰值聚类算法。该算法利用蝙蝠算法较强的寻优能力,寻找合适的截断距离取值,同时对蝙蝠算法的速度更新公式加入了自适应惯性权重来加强全局搜索能力。该算法选择多种数据集进行了实验仿真,并与其他同类算法进行对比。经过对比验证,结合蝙蝠算法改进的密度峰值聚类算法在聚类准确率上要明显优于其他算法。

Density peak clustering algorithm has the characteristics of fast clustering speed,simple implementation and fewer parameters,but the cutoff distance of the algorithm needs manual intervention,and the selection of the parameter has a greater impact on the results of the algorithm. To overcome this defect, an improved density peaks clustering algorithm combining bat algorithm is proposed. The algorithm used the bat algorithm′s stronger optimization capability to find the cutoff distance. And the adaptive inertia weight was added to the bat algorithm speed update formula to strengthen the global searching ability. The algorithm selected multiple data sets for experimental simulation and compared with other similar algorithms. Through comparison,improved density peaks clustering algorithm combining bat algorithm was more precise than other algorithms in clustering accuracy.