Which density-based clustering method defines groups as clusters and treats outliers as noise?

Which density-based clustering method defines groups as clusters and treats outliers as noise?

• A. SNN (Shared Nearest Neighbors)
• B. Density-based Clustering
• C. Agglomerative Clustering
• D. DBSCAN (Density-Based Spatial Clustering of Applications with Noise)

Answer: (D)

Density-based clustering treats clusters as dense regions of points and marks points outside these regions as noise. DBSCAN is the classic example: it uses two parameters, a radius and a minimum point count. If a point has at least the minimum number of points within its radius, it’s a core point; clusters grow by chaining together core points that are reachable through such density, while any point that isn’t density-reachable from a cluster is labeled as noise. This is why outliers are treated as noise rather than being assigned to a cluster. Other methods don’t inherently make this explicit: shared-nearest-neighbor approaches focus on neighborhood similarity to form groups but don’t inherently designate outliers as noise in the same density-based sense; hierarchical agglomerative clustering builds clusters based on linkage criteria and doesn’t independently tag isolated points as noise.

Density-based clustering treats clusters as dense regions of points and marks points outside these regions as noise. DBSCAN is the classic example: it uses two parameters, a radius and a minimum point count. If a point has at least the minimum number of points within its radius, it’s a core point; clusters grow by chaining together core points that are reachable through such density, while any point that isn’t density-reachable from a cluster is labeled as noise. This is why outliers are treated as noise rather than being assigned to a cluster.

Other methods don’t inherently make this explicit: shared-nearest-neighbor approaches focus on neighborhood similarity to form groups but don’t inherently designate outliers as noise in the same density-based sense; hierarchical agglomerative clustering builds clusters based on linkage criteria and doesn’t independently tag isolated points as noise.