STING 和 OPTICS 的区别

在数据集中识别相似数据组的方法称为聚类。每个组中的实体与该组中的实体比其他组中的实体更相似。

STING (统计信息网格聚类算法)和OPTICS (Ordering Point To identify Clustering Structure Clustering Algorithm)是无监督学习中使用的聚类算法。它们是机器学习技术，用于根据给定的输入数据点的属性将其划分为集群或组。 STING是基于网格的聚类算法，而OPTICS是基于密度的聚类算法。聚类

这些是 STING和OPTICS之间的一些区别。

S.No.	STING	OPTICS
1.	STING is abbreviation for Statistical Information Grid	OPTICS is abbreviaiton for Ordering Point To Identify Clustering Structure
2.	It is grib based clustering algorithm	It is density based clustering algorithm
3.	It concerns not with data points but with the value space that surrounds the data points.	It searches the data space for areas of varied density data points in the data space.
4.	It uses multi-dimensional grid data structure that quantizes space into a finite number of cells.	It is an extension to Density Based spatial clustering of applications with noise.
5.	The following are the properties of STING clustering algorithm:    Spatial area is divided into rectangular cells. Several level of cells at different levels of resolution. High level cell is partitioned into several low level cells. Statistically attributes are stored in cell for instance Mean, Maximum, Minimum are some of the statistical measures which are used. Statistical information is calculated for each cell and the types of distribution calculated are normal and exponential.	The following are the properties of OPTICS clustering algorithm:    It is an extension of DBSCAN, which takes the responsibility of parameters that can lead to discovery of unacceptable clusters. Core distance is the smallest point tha make a point core. Two important parameters are required for OPTICS: epsilon(“eps) and minimum points(“MinPts). The parameter eps defines the readius of neighborhood around a point P. The parameter MinPts is the minimum no. of neighbors within “eps”radius. Density = No. of points within a specified radius r(eps)
6.	It has relatively less computational complexity.	It has relatively more computational complexity.

STING 算法：

1. 首先确定一个层。
2. 对于该层的每个单元格，我们计算该单元格与查询相关的概率的置信区间(或估计范围)。
3. 根据上面的区间计算，我们将单元格标记为相关或不相关。
4. 如果这是底层，则结束该过程。
5. 我们将层次结构向下一层。对于形成更高级别层的相关单元格的那些级别，请转到步骤 2。

STING 层次结构图：

光学算法：
点 P 的核心距离是使 P 的邻域至少有 minPts 个点的最小距离。
p 到 q1 的可达距离是核心距离 (ε’)。
p 到 q2 的可达距离是 p 和 q2 之间的欧几里德距离。