In this report, we implement and compare DBSCAN in four different methods. These methods include naive, K-D Tree, Matrix and Triangle. The triangle method use the triangle inequality to divide the partitions. It use MapReduce and GraphFrame in spark to parallel compute the results. So it has the best performance in the large dataset and clusters.
- dbscan_naive.ipynb: The naive DBSCAN
- dbscan_kdtree.ipynb: The DBSCAN optimized by the kdtree
- dbscan_matrix.ipynb: The DBSCAN optimized by the matrix
- dbscan_triangle.ipynb: The DBSCAN optimized by the triangle inequality
use the remote environment in Azure: Environment configuration details
Wu hao
Chen zhizhen
Ji minghao
Huang yitian