Pollution Forecasting

Domain: Industry & Transport. ​

Project undertaken as part of a professionalization contract as a Data Scientist - Environmental Intelligence at Numtech, a company specialized in atmospheric modeling.

The goal was to strengthen environmental forecasting tools by simulating sensors, modeling pollution, and measuring impacts on air quality and health.

❓ Problem

Cities and industrial areas face difficult-to-detect and model diffuse pollution. Traditional systems are costly or not easily adaptable.
It was necessary to design a reliable and flexible predictive system, including:

• simulated sensors,

• algorithms for olfactory signal analysis,

• and models for predicting concentrations.

🛠️ Implemented Solution

• Sensor emulation to simulate real pollution environments.

• Olfactory signal processing: smoothing, advanced cleaning, filtering.

• Regression models to estimate the concentration of pollutants and odors.

• Tech watch on continuous monitoring systems.

• Statistical analysis and data visualization of results (curves, maps, time series).

• Popularized scientific documentation for public/private clients.

• Regular workshops with stakeholders to adjust business needs.

⚙️ Technical Stack

• Languages: Python · R · Shell (Linux)

• ML / stats libraries: scikit-learn · keras · tensorflow · xgboost · statsmodels

• Visualization: plotly · seaborn · matplotlib

• IDE / Environments: VS Code · Jupyter · Linux · PuTTY

• Data: weather, sensors, pollutants, time series

• Working language: French / English