Package com.here.platform.location.spark

Class DistributedClustering<GeolocatedItem>

java.lang.Object

com.here.platform.location.spark.DistributedClustering<GeolocatedItem>

All Implemented Interfaces:

Serializable, scala.Serializable

public class DistributedClustering<GeolocatedItem>
extends Object
implements scala.Serializable

Performs distributed clustering on a given collection of GeolocatedItem events.

The clustering is performed using a DBSCAN algorithm on each tile with a buffer zone of specified size. The clusters whose center lies on the buffer zone are rejected (as they are included in the output for a different tile).

Limitations:

- Due to the distribution scheme, this implementation returns incorrect results if there are clusters larger than the buffer zone. - DBSCAN identifies clusters by density, so the algorithm works best if all clusters have approximately the same spatial density.

param: neighborhoodRadiusInMeters The radius within which the algorithm searches for the neighboring events param: minNeighbors Minimal number of neighbors required to add the event to a cluster param: partitionBufferZoneInMeters Width of the zone that extends around each tile in order to support clusters spanning across the tile borders The buffer zone should be set to a larger size than the expected size of clusters param: partitionTileLevel The level of tiles used to distribute the clustering

See Also:

• Serialized Form

Constructor Summary

Constructors

Constructor	Description
DistributedClustering(double neighborhoodRadiusInMeters, int minNeighbors, double partitionBufferZoneInMeters, HereTileLevel partitionTileLevel, GeoCoordinateOperations<GeolocatedItem> evidence$2)

Method Summary

Modifier and Type	Method	Description
org.apache.spark.rdd.RDD<Cluster<GeolocatedItem>>	apply(org.apache.spark.rdd.RDD<GeolocatedItem> events, scala.reflect.ClassTag<GeolocatedItem> classTag)	Apply clustering to the given events.
scala.Tuple2<TileId,scala.collection.Iterable<Cluster<GeolocatedItem>>>	clusterEventsOnTile(scala.Tuple2<TileId,scala.collection.Iterable<GeolocatedItem>> tile)
final scala.collection.immutable.List<GeolocatedItem>	expandCluster(scala.collection.immutable.List<GeolocatedItem> currentCluster, boolean[] visited, scala.Tuple2<GeolocatedItem,Object>[] events, scala.collection.immutable.List<Object> toVisit)
scala.collection.Iterable<TileId>	getExtendedTileKeysForPoint(GeolocatedItem coords)
scala.collection.Iterable<Object>	getNeighboringIndices(scala.Tuple2<GeolocatedItem,Object>[] events, int i)
static <GC> scala.collection.Seq<Class<?>>	kryoClasses()
scala.collection.Iterable<Cluster<GeolocatedItem>>	removeClustersOnTileExtension(scala.Option<org.apache.spark.util.LongAccumulator> clustersCount, scala.Tuple2<TileId,scala.collection.Iterable<Cluster<GeolocatedItem>>> tile)

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Details

DistributedClustering

public DistributedClustering(double neighborhoodRadiusInMeters,
 int minNeighbors,
 double partitionBufferZoneInMeters,
 HereTileLevel partitionTileLevel,
 GeoCoordinateOperations<GeolocatedItem> evidence$2)

Method Details

kryoClasses

public static <GC> scala.collection.Seq<Class<?>> kryoClasses()

apply

public org.apache.spark.rdd.RDD<Cluster<GeolocatedItem>> apply(org.apache.spark.rdd.RDD<GeolocatedItem> events,
 scala.reflect.ClassTag<GeolocatedItem> classTag)

Apply clustering to the given events.

getExtendedTileKeysForPoint

public scala.collection.Iterable<TileId> getExtendedTileKeysForPoint(GeolocatedItem coords)

getNeighboringIndices

public scala.collection.Iterable<Object> getNeighboringIndices(scala.Tuple2<GeolocatedItem,Object>[] events,
 int i)

expandCluster

public final scala.collection.immutable.List<GeolocatedItem> expandCluster(scala.collection.immutable.List<GeolocatedItem> currentCluster,
 boolean[] visited,
 scala.Tuple2<GeolocatedItem,Object>[] events,
 scala.collection.immutable.List<Object> toVisit)

clusterEventsOnTile

public scala.Tuple2<TileId,scala.collection.Iterable<Cluster<GeolocatedItem>>> clusterEventsOnTile(scala.Tuple2<TileId,scala.collection.Iterable<GeolocatedItem>> tile)

removeClustersOnTileExtension

public scala.collection.Iterable<Cluster<GeolocatedItem>> removeClustersOnTileExtension(scala.Option<org.apache.spark.util.LongAccumulator> clustersCount,
 scala.Tuple2<TileId,scala.collection.Iterable<Cluster<GeolocatedItem>>> tile)