This repo is a recurrence of dynamic DBSCAN for paper: Dynamic Density Based Clustering with some optimization on the indexing, calculation of core points and the linkage of core points/core cells. The related data frames are based on the paper: DBSCAN Revisited: Mis-Claim, Un-Fixability, and Approximation and the reoccurence project is QuadForest. This repo has finished building and is under testing now. Please inform me when errors occurs, I will help to solve that.
This repo is based on Spark Streaming for dynamic data.
I have finished all algorithms and it now need to be tested: on performance and accuracy.
This implementation is based on the QuadForest for accelerating the process of maintaining the core points/cells. And if you want to use this repo as a mean of clustering dynamic and static data.
There are some deprecate files in src folder. They will be removed on next version.
From init file, this repo can achieve 640734/684017ms = 0.93 pts/ms (Testing evironment: Intel i7-8700 @3.2GHz, 24GB RAM, JDK-1.8, Scala-2.11, Win10Pro)
For testing, the command nc -l -p (port) to push data into processing stream.
Others(Theory, Implementation Mertics and Result) will given once I finish the test on this implementation.
Gan, Junhao, and Yufei Tao. "Dynamic density based clustering." Proceedings of the 2017 ACM International Conference on Management of Data. ACM, 2017.