Skip to content

Instantly share code, notes, and snippets.

@markmandel
Last active December 20, 2015 08:49
Show Gist options
  • Save markmandel/6102918 to your computer and use it in GitHub Desktop.
Save markmandel/6102918 to your computer and use it in GitHub Desktop.
First pass at a reference for the rec4j.properties file properties, and what they do.

reco4j.properties

This is the file that defines how reco4j builds recommendations between items in your neo4j database, when using reco4j as a plugin to neo4j.

Configuration Options

Property Name Default Description
recommenderType 1 (Collaborative Filtering) The type of recommendation engine to use.
1 Collaborative Filtering,
2 Matrix Factorization
DistanceAlgorithm 0 (Euclidean) The distance/similarity algorithm to use between items.
0 Euclidean
1 Euclidean Normalized
2 Jaccard
3 Cosine
nodeType type The property name to look for to determine if the node is a user or item type.
userType User The value of the nodeType property that marks this node as the type User
itemType Movie The value of the nodeType property that marks this node as the type Item
userIdentifier userId The node property to uniquely identify a node of User type
itemIdentifier itemId The node property to uniquely identify a node of Item type
KValue 25 The number of nearest neighbours to use when calculating recommendations (???? Is this right? Not sure what this means)
rankEdgeIdentifier rated The Edge type to use for rankings between an Item and a User (??? Is this right / could be explained better?)
RankValueIdentifier RankValue The Rank Value (??? What value should this have? Float/Int? Range between numbers? Also seems to be used on testRankEdgeIdentifier, not sure what this either )
testRankEdgeIdentifier ratedTest (??? I have no idea what this is)
recalculateSimilarity false Store the similarity value for future reload (??? Is this right?)

Examples

TODO: Include examples with explanations of what the configuration means

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment