In this paper, we propose a novel clustering technique that uses the simple idea of creating a graph on the data points based on nearest neighbors and identifying clusters by finding it’s connected components. The algorithm forms the graph based on a border detection and an outlier detection technique. We also propose a novel outlier detection technique that is suitable for our implementation. We compare our method against state-of-the-art clustering techniques and perform experiments to analyze the effects of various aspects of a dataset on it.

中文翻译：

---

选择性最近邻聚类

在本文中，我们提出了一种新颖的聚类技术，该技术使用基于最近邻居在数据点上创建图并通过查找其连接组件来识别集群的简单想法。该算法基于边界检测和异常值检测技术形成图形。我们还提出了一种适合我们实现的新颖异常值检测技术。我们将我们的方法与最先进的聚类技术进行比较，并进行实验以分析数据集的各个方面对其的影响。