Spark_DBSCAN

Spark_DBSCAN is a scala open source library of clustering algorithms implemented for Apache Spark

Installation

Download the jar file located here and copy it to your library

Example

import org.aizook.scala.clustering.Spark_DBSCAN.Dbscan
val eps = 3.0
val minPts = 5
val data = sc.textFile("data.csv").map(_.split(",")).map(p => (p(0).trim.toDouble, p(1).trim.toDouble)).zipWithUniqueId().map(x => (x._2,x._1)).cache
val cluster:Dbscan = new Dbscan(eps,minPts,data)
cluster.predict((data.count+1,(48.3,33.1)))

Should return -1 when the tested point fits noise or the cluster id otherwise

Release Note

0.0.1:

New features:

• DBSCAN Algorithm

0.0.2:

New features:

• Predict which clusters fit a list of points contained in a RDD

Bug fixing:

• [DBSCAN-0001] Predict function returns an exception when the tested point does not belong to a cluster
• luborliu/patch-2 Issue of the algorithm implementation besides the small condition
• luborliu/patch-3 Improve the efficiency of the code