This project implements an unsupervised approach to detect potential fraud in graph-structured data. It combines various graph analysis techniques and machine learning algorithms to identify anomalous patterns without relying on labeled training data.
Fraud detection in graph data presents unique challenges and opportunities. This project leverages the structural properties of graphs to identify potential fraudulent activities using unsupervised learning techniques.
- Graph construction from transaction data
- Feature engineering based on graph properties
- Community detection to identify unusual clusters
- Anomaly detection using Isolation Forest
- Graph visualization for intuitive understanding of results
- Python 3.7+
- NetworkX
- Pandas
- NumPy
- Scikit-learn
- Matplotlib
- python-louvain (for community detection)
- Clone this repository: