liveBook · Manning

Inferred network

As explained in Graph Algorithms for Data Science

An inferred network is a graph constructed by deriving relationships from observed graph patterns rather than storing those relationships directly.

An inferred network creates direct graph relationships from an indirect pattern already present in a graph. Rather than treating the derived relationship as original data, it summarizes what the observed pattern means for a particular analysis. The relationship type, direction, and properties must reflect domain assumptions. section 2.3.3

A retweet graph provides a practical example. Users publish tweets, and other users retweet them; this path can be translated into a direct AMPLIFY relationship between the relevant users. Under the assumption that retweets are positive interactions, AMPLIFY represents promotion of another user's content and the increased reach it receives. section 2.3.3 section 6.0

Listing 6.1 Describing the translation of the indirect retweet pattern to a direct AMPLIFY relationship
MATCH (s:User)-[:PUBLISH]->()-[:RETWEETS]->()<-[:PUBLISH]-(t:User)
CREATE (s)-[:AMPLIFY]->(t);

This Cypher pattern matches an indirect retweet path and creates a direct AMPLIFY relationship, producing a user-to-user projection from tweet-level activity. section 6.0

An inferred AMPLIFY network can retain useful analytical meaning:

  • It is directed when the direction of promotion is meaningful. section 2.3.3
  • It is weighted when the relationship stores the number of retweets between users; one retweet gives weight 1, while repeated retweets increase it. section 2.3.3 section 6.0
  • It forms a monopartite network containing User nodes and AMPLIFY relationships, removing Tweet nodes from the projected view. section 6.0
  • It supports influence and community analysis, such as finding users with highly sharable content or users who mutually promote content. section 2.3.3 section 6.0

Projecting retweets into amplification

Start with User–PUBLISH–Tweet and RETWEETS relationships. Match each retweet pattern, create a direct AMPLIFY relationship between the associated users, and count repeated retweets as a relationship property. PageRank on the resulting network can consider both retweet volume and who performed the retweet. section 6.0 section 2.3.3

How Graph Algorithms for Data Science covers it

How deep it goes
Defines the projection, direction, weighting, and a Cypher transformation, while deferring detailed projection caveats.
Examples
Uses Twitter users, tweets, retweets, AMPLIFY relationships, and a Cypher MATCH–CREATE projection.
What it assumes
Assumes a Twitter network is imported into Neo4j for the practical projection exercises.
What you can do after
Apply it via Cypher to project retweet patterns into a directed, weighted user network.
How it connects
Connects retweets, graph projection, monopartite networks, weighted relationships, PageRank, influencers, and communities.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Graph Algorithms for Data Science. The section links point to the parts of the book it draws on.