基于麻雀搜索算法改进的密度峰值聚类算法

Improved Density Peak Clustering Algorithm Based on Sparrow Search Algorithm

针对密度峰值聚类算法(Density Peaks Clustering Algorithm, DPC)用传统距离度量方式不能很好地反映数据分布,人为选取截断距离参数主观性较强等问题,设计了一种基于麻雀搜索算法改进的密度峰值聚类算法(Improved Density Peak Clustering Algorithm Based on Sparrow Search Algorithm, SSA-DPC)。该算法从两个方面进行改进:改变数据间的距离度量方式,用标准欧氏距离替代原算法中的欧氏距离;利用麻雀搜索算法(Sparrow Search Algorithm, SSA)较强的全局寻优能力,搜寻最佳截断距离值。通过对7个数据集进行仿真测试,证明SSA-DPC算法在3个评价指标上均优于其他聚类算法,提升了聚类性能,说明了算法的有效性。

Aiming at the problems that density peaks clustering algorithm (DPC) cannot well reflect the data distribution with traditional distance measurement, and the artificial selection of truncation dis-tance parameters is highly subjective, an improved density peak based on sparrow search algo-rithm was designed—Clustering algorithm (Improved Density Peak Clustering Algorithm Based on Sparrow Search Algorithm, SSA-DPC). The algorithm is improved from two aspects: change the distance measurement method between data, and replace the Euclidean distance in the original algorithm with the standard Euclidean distance;using the strong global optimization ability of the Sparrow Search Algorithm (SSA), the best cutoff distance value was searched. Through the simula-tion test of 7 data sets, it is proved that the SSA-DPC algorithm is superior to other clustering algo-rithms in 3 evaluation indicators, and the clustering performance is improved, which shows the effectiveness of the algorithm.