The commonly used density-based spatial clustering method (DBSCAN) connects contiguous regions with sufficiently large densities when processing datasets to efficiently discover clusters of different shapes and densities and outliers. However, the algorithm has the problem that radius of neighborhood (Eps) argument requires to be selected manually. For datasets with higher dimensionality and larger data volume, the selection of Eps parameters can be difficult thus leading to poor clustering quality. To solve the above problem, we propose a novel adaptive density-based spatial clustering of application with noise based on bird swarm optimization algorithm (BSA-DBSCAN). We use the global search capability of the bird swarm method to select the best Eps parameter neighborhood values. We can avoid manual intervention and realize adaptive parameter optimization in the clustering process. To further explore the clustering performance of BSA-DBSCAN method, we test the synthetic datasets and the real-world datasets respectively and perform images analysis on the clustering evaluation index values. The simulation experiments show that the improved method in this paper can reasonably search the Eps parameter value and can obtain the higher accuracy of clustering.

在处理数据集时，常用的基于密度的空间聚类方法（DBSCAN）将具有足够大密度的连续区域连接起来，以有效地发现具有不同形状，密度和异常值的聚类。但是，该算法具有需要手动选择邻域半径（Eps）参数的问题。对于具有较高维度和较大数据量的数据集，可能难以选择Eps参数，从而导致较差的聚类质量。为了解决上述问题，我们提出了一种基于鸟群优化算法（BSA-DBSCAN）的基于自适应密度的带噪声应用空间聚类的新方法。我们使用鸟群方法的全局搜索功能来选择最佳的Eps参数邻域值。在聚类过程中，可以避免人工干预，实现自适应参数优化。为了进一步探索BSA-DBSCAN方法的聚类性能，我们分别测试了合成数据集和真实数据集，并对聚类评估指标值进行了图像分析。仿真实验表明，本文提出的改进方法可以合理地搜索Eps参数值，并获得较高的聚类精度。