Using Graph Theory to Build a Recommendation Engine in Node js
summary
Claims a working Node js recommendation engine that replaces a slow SQL job by loading users listings and interactions (views favorites requests) into RAM as a graph, weighting edges (request=1 favorite=2 view=4), running a modified Dijkstra breadth‑first traversal to collect 100 candidates and sampling three to show. Implementation runs on a separate server regenerated daily and snapshotted to disk, ships as unit graph on GitHub, ties into the nodal API, is built in about a week and answers queries in <1ms.
AI-written summary · about
chapters (23)
- 0:00Intro
- 0:11About Me
- 1:35Problem Space
- 2:29Solution and Requirements
- 4:34First Solution Problems
- 5:42Introducing the Graph
- 7:00Graph-Based Solution Benefits
- 7:55What Technologies Available?
- 8:29Let's Build A Graph Ourselves!
- 9:36Building A Graph in JavaScript
- 10:44What do our Nodes and Edges need?
- 11:49Good Case for Inheritance!
- 12:46Base Unit class
- 13:28Edge More Complicated
- 14:37Re-Implement Basic Graph
- 15:02Add More Complexity
- 16:14Implementation (Setup)
- 20:29Weight your Edges!
- 20:41Use Edge#set Distance...
- 21:01Putting it all together...
- 22:48Final Implementation (Cont.)
- 24:15Conclusions
- 24:37Code and Related Projects
related talks
38mRecommendation Engine with Neo4j, Piece of Cake
46mYour own recomendation Engine with Neo4j AND Reco4PHP
01h 11mBuilding a Recommendation Engine with Neo4j and Python
21mUnlocking Real-Time News Recommendations: The Power of Graph Relationships in Neo4j
55mBuilding a Recommendation Engine with Machine Learning Techniques
01h 38mBuilding a recommendation engine with Python and Neo4j
26mGraph Databases will change your (freakin') life
51mBuilding a Meetup Recommendation Engine with Spring Data Neo4j