Searching for Research Fraud in OpenAlex with Graph Data Science
summary
Project ingests a 1.7 TB OpenAlex snapshot into Neo4j on GCP, building subgraphs from ~1.7 billion relationships and ~400 million author nodes using Airflow pipelines; queries run in minutes on a £20 machine. Features include min/max prior works, bibliographic coupling and Jaccard-style concept overlap; a ML classifier separates brokered from normal co-authorships. Results show ML performs well but a simple heuristic—first-paper with different-institution rule—is almost as effective. Key stack: OpenAlex + Neo4j on GCP, Airflow, graph analytics.
AI-written summary · about
chapters (23)
- 0:00Introduction
- 0:39Coauthorships
- 1:49Referee Accounts
- 3:28What are Paper Mills
- 5:41Detecting Paper Fraud
- 9:00About OpenAlex
- 10:14Data Engineering
- 11:28Hercules
- 12:41Acceptance
- 13:35Linking Concepts
- 14:15Twitter ID
- 15:06Author ID
- 15:37ingestion process
- 16:01code helps
- 16:53subgraph
- 17:36labels
- 18:34similarity
- 20:07relationships
- 20:37exploring API
- 21:20data analysis
- 24:34results
- 26:39machine learning
- 27:18summary
related talks
01h 41mGraph Intelligence: Enhance Reasoning and Retrieval Using Graph Analytics
30mPerformant GraphRAG Improving Neo4j to Drive Autonomous AI ...
42mFighting fraud with graph databases
39mEnhancing RAG with Neo4j Knowledge Graph
15mAgentic GraphRAG: Simplifying Retrieval Across Structured & Unstructured Data
01h 00mGraph Mastery to AI
37mBuilding a Scientific Taxonomy at Scale
46mFraud Detection at Floa