Boundary point is a kind of data object with special significance in data set. They are located in the edge area of density-based clusters. Boundary point processing is of great significance in data mining. They represent a class of individuals with unclear ownership. If we simply rely on a certain method to classify them into a specific cluster, the effect is often counterproductive. Boundary points are different from isolated points and noise points. Outliers are a kind of objects in a minority position in statistics, and noise points are objects that interfere with statistics or deviate from a certain distribution. They are usually located in the low-density area of data space, and the boundary points are different. They are a kind of data objects on the edge of high-density areas in data space, with high-density areas on one side and relatively low-density areas on the other.

The research of clustering technology is a hot spot in recent years, and many clustering algorithms have been proposed, but there are not many discussions on clustering boundary patterns. The boundary point of clustering refers to a class of data objects located at the edge of high-density clustering, representing a class of individual objects floating between two or more categories, whose ownership is not clear, and often has two or more clustering characteristics. The study of boundary points has important application value.