Label Propagation Algorithm

Introduction

The Label Propagation Algorithm (LPA) is a fast and efficient method for detecting communities within a graph.

Output

The algorithm outputs the following fields:

Field Name	Description
id	The ID of the vertex (node)
communityId	A community ID assigned to the node.

Using Cypher

Query Example

CALL algo.labelPropagation({
    labels: ['User'],
    relationshipTypes: ['LINK'],
    relationshipWeightProperty: 'weight',
    labelProperty: 'tag'
    }) YIELD id, communityId
RETURN id, communityId

Parameters

Name	Description	Required
labels	Specifies the node labels to include in the algorithm	Yes
relationshipTypes	Specifies the types of relationships to include	Yes
relationshipWeightProperty	Defines the property name representing edge weights	No
labelProperty	The property used to assign an initial community ID	No
pinLabel	Determines if the initial label of vertices remains fixed	No

Exporting Query Results

For large datasets, export the results to object storage:

EXPORT TO 's3://my_bucket/target_dir'
PROPERTIES {
  catalog: 'puppygraph_catalog_name'
}
CALL algo.labelPropagation({
    labels: ['User'],
    relationshipTypes: ['LINK'],
    relationshipWeightProperty: 'weight'
    }) YIELD id, communityId
RETURN id, communityId

Using Gremlin

Query Example

graph.compute().program(
  LabelPropagationProgram.build()
    .maxIteration(20)
    .vertices("User")
    .edges("LINK")
    .edgeProperties("amount")
    .setParams("weight", "amount")
    .setParams("labelProperty", "tag")
    .setParams("pinLabel", true)
    .create()
).submit().get().toList()

Parameters

Method	Description
maxIteration(int n)	Sets the maximum iteration count, default is 30
vertices(String... names)	Specifies the vertex labels to include
edges(String... names)	Specifies the edge labels to include
edgeProperties(String... names)	Sets the property name for edge weights
setParams(Object... args)	Adds additional parameters using alternating keys and values

Additional Parameters

Key	Description	Default Value
weight	Specifies edge weight property or a constant number	1.0
labelProperty	Property name for assigning an initial community ID	""
pinLabel	Determines if the initial label of vertices remains unchanged	false

Exporting Query Results

For large datasets, use the save() method to export the results.

graph.compute().program(
    LabelPropagationProgram.build()
            .maxIteration(20)
            .vertices("User")
            .edges("LINK")
            .edgeProperties("amount")
            .setParam("weight", "amount")
            .create()
).submit().get().save([
  "exportTo": "s3://my_bucket/target_dir",
  "catalog": "puppygraph_catalog_name"
])