# Graph for fraud detection

[Grab](https://yomu.fyi/company/grab) · Min Chen · Nov 24, 2022

**Type:** Problem & solution

## Summary

Rapid expansion across multiple business verticals introduced diverse, evolving fraud patterns at Grab, making traditional rule engines and decision trees inadequate due to their heavy reliance on extensive feature engineering and labeled data. To address cold starts and detect unknown fraud, the engineering team implemented a semi-supervised Relational Graph Convolutional Network (RGCN) across millions of connected ecosystem entities. The model propagates information along neighborhood nodes through graph convolutional layers to produce node embeddings and output fraud probabilities. Trained on graphs where only a small percentage of nodes are labeled, the RGCN achieved an AUROC close to 1 and distinctly separated fraudulent embeddings from genuine ones. The approach also offers high explainability by surfacing dense clusters of shared physical devices and addresses ongoing challenges in real-time prediction and noisy connections.

## Context

Rapid business expansion into new verticals created diverse, evolving fraud patterns where traditional rule engines and decision tree models faced cold start problems and struggled to detect unknown fraud without extensive labels and feature engineering.

## Approach / What changed

Implemented a semi-supervised Relational Graph Convolutional Network (RGCN) model on graphs with millions of nodes and edges to capture structural correlations across shared entities with minimal labeled data.

## Takeaways

- Using fewer than three convolutional layers prevents over-smoothing node features in the Relational Graph Convolutional Network.
- Semi-supervised RGCN models can achieve high classification accuracy with AUROC near 1 even when only a small percentage of nodes contain labels.
- Graph neural network predictions provide high explainability because fraudulent accounts naturally manifest as dense clusters sharing physical properties like hardware devices.

**Tags:** [Architecture](https://yomu.fyi/topic/architecture), [Machine Learning](https://yomu.fyi/topic/machine-learning), [Scalability](https://yomu.fyi/topic/scalability)

[Read original post](https://engineering.grab.com/graph-for-fraud-detection)