How koharani-2011/graph-fraud-analysis works
Graph Fraud Analytics combines a browser-based upload/results dashboard, a FastAPI analysis endpoint, and graph-model training and evaluation code. The demonstrated upload path accepts CSV/Excel, sends it through the frontend API boundary, constructs graph data in the backend service, runs EvolveGCN inference, and returns summary counts and a time series for display. Separate temporal training code prepares time-step snapshots and evaluates future steps; custom GAT and benchmark models provide documented alternatives. The frontend API’s HTTP target is not shown, so that integration is marked unverified.
Main entry points: App.jsx Dashboard app · UploadSection.jsx Dataset upload · train_temporal.py Temporal training · main.py FastAPI endpoint
Components 17
Analysis dashboard
- Dataset upload
frontend/src/components/UploadSection.jsx - Frontend API
frontend/src/api.jsx - Results dashboard
frontend/src/components/DashboardResults.jsx - Dashboard app
frontend/src/App.jsx - Graph logo
frontend/src/components/Logo.jsx
Inference service
- FastAPI endpoint
backend/main.py - File analysis
backend/services.py
Models and training
- EvolveGCN inference
models/evolvegcn.py - Temporal snapshots
train_temporal.py - Temporal training
train_temporal.py - Custom GAT
custom_gat/model.py - Attention layer
custom_gat/layer.py - GNN benchmarks
models/models.py - Training runner
train.py
Evaluation outputs
- Model evaluation
evaluate_and_plot.py - Metric plotting
plot_results.py
Other components
- Analyst
Connections 18
- Analyst → Dataset upload: selects file
frontend/src/components/UploadSection.jsx - Dataset upload → Frontend API: calls analysis
frontend/src/components/UploadSection.jsx - Frontend API → FastAPI endpoint: submits file
- FastAPI endpoint → File analysis: calls analysis
backend/main.py - File analysis → EvolveGCN inference: runs inference
backend/services.py - FastAPI endpoint → Frontend API: returns results
- Dashboard app → Dataset upload: renders
frontend/src/App.jsx - Dashboard app → Results dashboard: renders
frontend/src/App.jsx - Dashboard app → Graph logo: renders
frontend/src/App.jsx - Dataset upload → Dashboard app: sets results
frontend/src/App.jsx - Results dashboard → Analyst: shows summary
frontend/src/components/DashboardResults.jsx - Temporal training → Temporal snapshots: uses snapshots
train_temporal.py - Temporal training → EvolveGCN inference: trains or evaluates
train_temporal.py - Temporal training → Custom GAT: trains baseline
train_temporal.py - Custom GAT → Attention layer: composes layers
custom_gat/model.py - Training runner → Custom GAT: imports model
train.py - Model evaluation → EvolveGCN inference: uses model
evaluate_and_plot.py - Model evaluation → Metric plotting: produces plots