Test ofnatural clusters via s-dbscan a self tuning version of DBSCAN

Learning notes , For reference only , If there is a mistake, it must be corrected
Journal：Knowledge-Based Systems
Year：2022
Keywords：Clustering,Natural cluster,Distance,Density,Neighbors

---

Detection ofnatural clusters via S-DBSCAN a Self-tuning version of DBSCAN

abstract

Density based clustering algorithm has a great impact on a wide range of applications . As the types and quantities of data increase , Its scale and various internal organizations are rising , Nonparametric unsupervised programs are becoming increasingly important in understanding datasets . This paper proposes a new clustering algorithm under the background of knowledge discovery S-DBSCAN. S-DBSCAN It belongs to the clustering algorithm based on connectivity ( Such as DBSCAN), But because it works in differential mode , So there are obvious differences and advantages . It is through a very simple layering process , Mixed distance 、k-nearest And density peak . Its purpose is to divide the existing data into clusters , Until you can no longer cluster . The information provided allows users to intuitively infer different sets of natural partitions in clusters of different sizes . S-DBSCAN By applying its algorithm core （S-DBSCANCORE） And input parameters , Scan data in an orderly manner . Given a set of data points in a certain space ,S-DBSCANCORE Group closely linked sets of data points . Those data points whose nearest neighbor density difference is too large will be detected and marked as boundaries , Others are not visited . S-DBSCAN Embeddedness can adjust itself （ Almost automatic ）, Unlike many existing algorithms that rely on a global density threshold . We use various shapes and densities