Exploring Neo4j and Graph Data Science for Fraud Detection

This repository contains the notebook for reproducing the fraud detection analysis covered in this blog series. To summarize, this analysis uses Neo4j and Graph Data Science (GDS) to explore an anonymized data sample from a Peer-to-Peer (P2P) payment platform. The notebook (fraud-detection-demo-with-p2p) is split up into the following sections, mirroring the blog series, to cover various stages of the graph data science workflow:

• Part 1: Exploring Connected Fraud Data
• Part 2: Resolving Fraud Communities using Entity Resolution and Community Detection
• Part 3: Recommending Suspicious Accounts With Centrality & Node Similarity
• Part 4: Predicting Fraud Risk Accounts with Machine Learning

Getting the Data

To run the notebook you will need a copy of the dataset which is available in the form of a neo4j dump file in this folder. The folder also contains a readme with more details on the dataset and directions for how to load the data into neo4j if you are unfamiliar with the process. The folder contains the ODC-BY license for the dataset as well (separate from the license in this repository).

Subdirectories & Variants

There are a couple subdirectories containing variants of the fraud-detection-demo-with-p2p notebook for various purposes:

• ./gds-v1.8 contains a GDS 1.8 compatible version of the notebook. The notebook in the primary directory is built for GDS 2.x and will not work on GDS 1.8 or earlier versions.
• ./abridged-demo contains a version of the notebook for a quick introductory demo. Some steps have been removed and altered to shorten the notebook and others steps have been summarized/condensed. This notebook is also built for GDS 2.x and not 1.x compatible.