Regulator

Advanced Market Collusion Detection: 93% accuracy with ML+LLM detection, real-time monitoring, and economic validation

Quickstart

Clone the repo and run the demo:

git clone https://github.com/bangyen/regulator.git
cd regulator
pip install -e .
pytest   # optional: run tests
python scripts/run_experiment.py --firms "random,titfortat" --steps 100

Or open in Colab: Colab Notebook.

For real-time monitoring, run the dashboard:

python dashboard/main.py
# Or use the CLI: regulator dashboard
# Visit http://localhost:5000

Development Workflow

This project uses uv for dependency management and just as a task runner.

• Initialize environment: just init
• Format code: just fmt
• Lint: just lint
• Type check: just type
• Run tests: just test
• Run all checks: just all

Results

Detection Method	Accuracy	Precision	Recall	F1 Score	ROC AUC
ML Detector	93.0%	85.0%	97.1%	90.7%	98.0%
LLM Detector	69.0%	49.2%	96.7%	65.2%	91.8%

Features

• Real-Time Dashboard — Professional monitoring interface with live metrics, charts, and violation tracking.
• ML+LLM Detection — 93% accuracy with sub-millisecond processing for real-time monitoring.
• Economic Validation — Built-in consistency checks and market dynamics validation.
• Chat Integration — Natural language communication monitoring with OpenAI-powered analysis.
• Enhanced Regulator — Graduated penalties, continuous risk scores, and market-aware monitoring.
• Leniency Programs — Strategic whistleblower dynamics and evidence-based reporting.

Repo Structure

regulator/
├── regulator_demo.ipynb  # Colab notebook
├── dashboard/            # Flask dashboard (run with dashboard/main.py)
├── scripts/              # Example run scripts
├── tests/                # Unit/integration tests
├── docs/                 # Images for README
└── src/                  # Core implementation
    └── regulator/        # Main package
        ├── agents/       # Market agents (firm_agents.py, leniency.py)
        ├── cartel/       # Market environment
        ├── detectors/    # Detection systems (under development)
        ├── episode_logging/ # Logger & episode runner
        ├── experiments/  # Experiment logic
        ├── monitoring/   # Placeholder for future metrics
        └── cli.py        # CLI entry point

Validation

• ✅ Overall test coverage of 92% (pytest)
• ✅ Reproducible seeds for experiments
• ✅ Benchmark scripts included

References

• Algorithms, Machine Learning, and Collusion - Comprehensive analysis of self-learning algorithms and collusive outcomes
• Deep learning for detecting bid rigging - CNN-based approach for flagging cartel participants using pairwise bidding interactions
• Algorithmic Collusion: A Critical Review - Critical assessment of pricing algorithms and collusion potential

License

This project is licensed under the MIT License.