TopicRank

Overview

TopicRank algorithm is one of commonly used algorithms for ranking topics by multiple dimensions.

Application Scenarios

This algorithm is applicable to rank hot topics. For example, it can be used to rank complaint topics obtained through a government hotline.

Parameter Description

**Table 1** TopicRank parameters
Name	Mandatory	Description	Type	Value Range	Default
sources	Yes	Vertex ID. You can specify multiple IDs in CSV format and separate them with commas (,).	String	Currently, a maximum of 100000 IDs are allowed.	-
actived_p	No	Initial weight of the source vertices	Double	The value ranges from 0 to 100000.	1
default_p	No	Initial weight of a non-source vertices	Double	The value ranges from 0 to 100000.	1
filtered	No	Whether to filter results	Boolean	The value can be true or false.	false
only_neighbors	No	Whether to display only the neighboring vertices of the sources	Boolean	The value can be true or false.	false
alpha	No	Weight coefficient	Real number	A real number between 0 and 1	0.85
convergence	No	Convergence	Real number	A real number between 0 and 1	0.00001
max_iterations	No	Maximum iterations	Positive integer	The value ranges from 1 to 2000.	1000
directed	No	Whether the edges are directed	Boolean	The value can be true or false.	true
num_thread	No	Number of threads	Positive integer	1-40	4

Example

Specify sources="20190110004349,20190129023326,20190107003294,20190129023391", filtered = true, only_neighbors=true, alpha=0.85, converage=0.00001, max_iterations=1000, directed=true, and label="Topic" to obtain the topic ranking result.

Parent topic: Algorithms

Previous topic: Filtered All Shortest Paths

Next topic: Filtered n-Paths