How DBSCAN clustering works? | AI ML tutorials by a Data Scientist | Thinking Neuron

Описание: https://thinkingneuron.com/data-scien...
DBSCAN stands for Density-Based Spatial Clustering of Applications with Noise.

Do not get intimidated by this algorithm’s name, the more it sounds complex the simpler it is actually! 🙂

1. Start with one point randomly, find out if it is a core point by checking the minimum number of points near to it by a distance of eps
2. If it is a core point, make it a cluster and move to the next unvisited point to repeat step-1
3. If the number of points within eps distance is fewer than MinPts then mark it as non-core point
4. If the number of points within eps distance is Zero, then mark that point as Noise.
5. Combine all those clusters together whose points are within eps distance. Also known as density connection or connected components. This starts a chain reaction. If cluster-1 and cluster-2 are connected and cluster-2 and cluster-3 are connected then cluster-1 and cluster-3 are also connected. Hence all these are combined to make one single cluster.