1. DBSCAN algorithm
DBSCAN algorithm, the full name is "Density-Based Spatial Clustering of Applications with Node", which is also "density-based clustering". This type of algorithm assumes that the clustering structure can be determined by the closeness of sample distribution, examines the continuity between samples from the perspective of sample density, and continuously expands clusters based on connectable samples to obtain the final clustering result.
The DBSCAN algorithm is a well-known density clustering algorithm. It uses neighborhood parameters (Distance, MinPts) to characterize the tightness of sample distribution. Before actually starting, you must first understand the following concepts:
Core objects: x1, x2, which are points that meet the conditions of neighborhood parameters (Distance, MinPts). Density direct: x2 is directly connected by density x1. Density reachable: x3 is reachable from x1 density. Density connected: x3 and x4 are density connected. For specific definitions, please refer to the book "Machine Learning". I will just briefly explain these concepts here.
2. Algorithm steps
1. Initialize the core object and record its adjacent points. Loop through each data point, record other data points <= Distance parameter (distance parameter) in the neighborhood of each data point, and determine whether the number of adjacent points that meet the distance parameter >= MinPts parameter (point number parameter), if so Add adds the data point to the collection of core objects. 2. Clustering. First select a core object in the data set as the "seed", and then start from this to find all its density-reachable sample sets to generate corresponding clusters until all core objects have been visited.
3. Code implementation effect
Original data graph:
main.cpp：