aniissaiidii/tunisian-scout-platform
Record type:
project
Creator:
ani
Host:
End-to-end ML platform for the Tunisian Scout Association to forecast activity participation. Includes Angular frontend, FastAPI ML backend (classification, regression, clustering, time series, LSTM), and Power BI dashboards. Production-ready with Docker, MLflow, Prometheus, and Grafana.
# Tunisian Scout ML Platform
End-to-end platform for the Tunisian Scout Association to forecast activity participation, segment scout units, and provide strategic dashboards.
## Project Overview
The association organizes hundreds of activities yearly across Tunisia, France, Spain, Italy, and America. This platform helps leaders estimate participant numbers, optimise budgets (21.14M TND), and monitor performance.
**Three integrated axes:**
- **Machine Learning** – Classification, regression, clustering, time series, LSTM, RL, NLP.
- **MLOps** – FastAPI, MLflow, Prometheus, Grafana, n8n, Docker.
- **Power BI** – Role-specific dashboards (Fawj Leader, Treasurer, Unit Leader, International).
## ML Models & Performance
| Module | Task | Best Model | Key Metric |
|--------|------|------------|-------------|
| Classification | Participation level (low/medium/high) | Random Forest | Macro F1 = 0.59 |
| Regression | Exact participant count | Random Forest Regressor | R² = 0.82, RMSE = 8.5 |
| Clustering | Unit segmentation | KMeans (k=2) | Silhouette = 0.69 |
| Time Series | Daily attendance forecast | SARIMA / LSTM | 98.9% accuracy |
| NLP | Text classification | TF‑IDF + Logistic Regression | – |
| Reinforcement Learning | Activity type per season | Q‑learning (ε‑greedy) | Cumulative reward |
| Deep Learning | Sequence forecasting | LSTM(50) + Dense(1) | RMSE |
## MLOps Stack
- **FastAPI** – Prediction API (`/predict`, `/retrain`, `/metrics`, `/health`)
- **MLflow** – Experiment tracking (12 runs, versioned models)
- **Prometheus** – Metrics scraping (9 alerts configured)
- **Grafana** – 11 real‑time monitoring panels
- **n8n** – Automated workflows (daily prediction, weekly retraining)
- **Docker Compose** – Orchestration of all services
## Power BI Command Center
Multi‑page report with role‑based views.
### Fawj Leader Dashboard
- Unit growth comparison (2023‑2025)
- Engagement ratio (15.26% – 17.68%)
- Event attendance by unit & activity type
- Spending effic …