Anomaly Detection for Water & Energy Consumption
SeafaringIT
StageSur site4 à 6 moisDate limite : 23 févr. 2026
Data Science & Machine LearningData Analysis / Data ScienceAnomaly detectionExplainabilityMLOps / DevOpsAPIs & Integrations
Description
- Goals:
- Build an intelligent anomaly detection system to identify water leaks and energy overconsumption
- Provide explainable insights into anomaly causes
- Prioritize interventions by impact and savings
- Student roles: Machine learning engineers, data scientists, backend engineers, MLOps specialists
- Expected outcomes: ML-powered anomaly detection with explainability, API exposure, model tracking, and impact analysis
- Key features:
- Dataset construction (open/synthetic) with feature engineering
- Multiple anomaly detection models (statistical + ML) with benchmarking
- Explainability layer (SHAP) and error analysis
- Alert API with scoring and prioritization
- Model tracking/versioning with MLflow
- Concept drift monitoring and performance tracking
- Business impact report (savings, environmental impact)
- Technologies: Python, scikit-learn, XGBoost, Isolation Forest, SHAP, MLflow, FastAPI, PostgreSQL, Elasticsearch, visualization tools