densitetbased

Density‐based clustering is a method for discovering clusters in spatial data that are defined by high density of points rather than geometric shape. The most well known approach is DBSCAN, which groups points that have at least a minimum number of neighbours inside a given radius. Points that do not meet these density criteria are classified as noise. Other algorithms, such as OPTICS and HDBSCAN, extend the basic idea by allowing variable neighbourhood sizes or producing hierarchical cluster structures.

The core concept relies on two parameters: ε (neighbourhood distance) and MinPts (minimum number of points required to form a dense region). For each point, the algorithm counts how many other points lie within ε. If this count is at least MinPts, the point is deemed a core point. A core point and all points within ε of it belong to the same cluster, and the process continues transitively through neighbouring core points. Border points, near the edge of a dense region, are linked to a core point but do not propagate the radius further.

Advantages of density‐based methods are their ability to find arbitrarily shaped clusters, independence from initial conditions, and robust handling of noise. They can discover clusters of varying sizes if parameter selection is appropriate. Limitations include sensitivity to the ε and MinPts parameters, difficulty scaling to very high dimensional data, and performance degradation when data are non‐uniformly distributed.

In practice, density‐based clustering is applied in spatial pattern analysis, image segmentation, anomaly detection, and geographic information systems. Parameter tuning is critical; automated heuristics or domain knowledge often guide the selection of ε and MinPts. Despite its challenges, density‐based clustering remains a foundational technique for exploratory data analysis and pattern recognition.