🏴‍☠️ Oasis Security – Crime Predictor

A predictive model for crime rates in France (per 100,000 inhabitants), fully structured for production use.
This repository demonstrates a ML project with clear structure, documentation, modeling pipeline, and deployment readiness.

🚀 Project Objectives

• Build a reliable predictive model using official French crime data (from data.gouv.fr).
• Establish a clean, reproducible MLOps pipeline:
  • Clear folder structure (data/, notebooks/, models/, etc.)
  • Serialized model for reuse
  • Training & inference scripts
  • Integrated CI/CD workflows

---

📁 Repository Structure

oasis-security/
├── .github/ # GitHub workflows (CI/CD)
├── data/ # Processed data files
├── docs/ # Documentation & dashboards
├── images/ # Visual assets & plots
├── models/
│ └── crime_predictor/
│ ├── src/ # Source code for model
│ ├── models/ # Serialized model (.pkl)
│ ├── mlruns/ # MLflow tracking data
│ ├── tests/ # Unit tests (optional)
│ └── requirements.txt # Dependencies for this model
├── notebooks/ # Exploration & analysis notebooks
├── pipeline/ # Scripts for automation
├── Dockerfile # Docker configuration
├── LICENSE # License
└── README.md # Project overview

---

📊 Usage

1. Create & activate a virtual environment: python3 -m venv .venv source .venv/bin/activate

2. Install dependencies: pip install -r models/crime_predictor/requirements.txt

3. Run training: python models/crime_predictor/src/train.py

4. Start prediction API: python models/crime_predictor/src/predict.py

📝 Contribution & CI/CD

This project is designed to be production ready with GitHub Actions workflows (tests & model builds). Contributions welcome 🌟

🛠️ Tech Stack

Core: Python 3.13, scikit-learn, joblib Future: FastAPI, MLflow, Docker, GitHub Actions Data: data.gouv.fr (police/gendarmerie 2016-2025)

📜 License

MIT License

📝 Author

Frédéric Tellier – Data Scientist LinkedIn : https://www.linkedin.com/in/fr%C3%A9d%C3%A9ric-tellier-8a9170283/
| Portfolio : https://github.com/Dreipfelt/