How aakritiiii16/aml-network-analyzer works
This is an AML analyst platform: transaction IDs enter through the Streamlit portal, are scored by the FastAPI service, and return a risk probability with contribution values for display. A separate API action produces a sampled performance audit. The documented modeling workflow constructs and temporally splits the transaction graph, creates embeddings, combines them with local features, and classifies transactions. README descriptions support these modeling-stage links; the notebooks are present in the tree but were not sampled, so their internal implementations and any unshown connections are not asserted.
Main entry points: app.py Streamlit Portal · main.py FastAPI Service · 1_make_graph.ipynb Transaction Graph · 2_embeddings.ipynb GNN Embeddings
Components 16
Graph Preparation
- Bitcoin Transactions
- Transaction Graph
notebooks/1_make_graph.ipynb - Temporal Split
notebooks/1_make_graph.ipynb
Modeling
- GNN Embeddings
notebooks/2_embeddings.ipynb - Hybrid Graph Model
notebooks/2_embeddings.ipynb - Feature Merge
notebooks/2_embeddings.ipynb - LightGBM Classifier
notebooks/3_train_LGBM.ipynb - SHAP Explanations
notebooks/4_SHAP_values.ipynb
Scoring And Monitoring
- FastAPI Service
api/main.py - Transaction Scoring
api/main.py - Performance Audit
api/main.py - Evidently Metrics
api/main.py
Analyst Portal
- Streamlit Portal
frontend/app.py - Risk And Explanation View
frontend/app.py - Monitoring Dashboard
frontend/app.py
Other components
- AML Analyst
Connections 17
- Bitcoin Transactions → Transaction Graph: constructs
README.md - Transaction Graph → Temporal Split: splits chronologically
README.md - Temporal Split → Hybrid Graph Model: trains leak-proof
README.md - Hybrid Graph Model → GNN Embeddings: generates embeddings
README.md - GNN Embeddings → Feature Merge: merges with local features
README.md - Feature Merge → LightGBM Classifier: classifies
README.md - LightGBM Classifier → SHAP Explanations: explains predictions
README.md - AML Analyst → Streamlit Portal: enters transaction ID
README.md - Streamlit Portal → FastAPI Service: POST /predict
frontend/app.py - FastAPI Service → Transaction Scoring: routes prediction
api/main.py - Transaction Scoring → Risk And Explanation View: returns probability and shifts
api/main.py - Streamlit Portal → Risk And Explanation View: renders result
frontend/app.py - AML Analyst → Monitoring Dashboard: requests audit
README.md - Streamlit Portal → FastAPI Service: POST /performance
frontend/app.py - FastAPI Service → Performance Audit: routes audit
api/main.py - Performance Audit → Evidently Metrics: builds classification report
api/main.py - Evidently Metrics → Monitoring Dashboard: returns HTML report
api/main.py