:no_entry:[No more maintained]

DBSCAN

DBSCAN (Density-Based Spatial Clustering of Applications with Noise) is a data clustering algorithm proposed by Martin Ester, Hans-Peter Kriegel, Jörg Sander and Xiaowei Xu in 1996. The algorithm had implemented with pseudocode described in wiki, but it is not optimised.  

Example

You can test this DBSCAN algorithm with example code(main.cpp) & sample data(benchmark_hepta.dat).

Results

Clustering performance was tesed via clustering-benchmark dataset and real-world dataset.

Build

$ g++ main.cpp dbscan.cpp -o dbscan

benchmark dataset

Artificial dataset was used to test clustering algorithm which can be obtained here. Following parameters were used:

1. Minimum number of points: 4
2. Epsilon: 0.75  

Source: Hepta (Total number of cluster: 7)