The Weather Shield Breda project aimed to enhance urban safety by predicting the severity of road incidents in Breda based on weather conditions and environmental factors. The initiative combined diverse datasets—including traffic incident logs, weather data (precipitation, temperature, wind speed), driver behavior records, and road/greenery information—to model how weather influences accident risks across the city.

Together with my team I applied machine learning methods such as K-Nearest Neighbors (KNN), Random Forest, and LSTM networks to classify incident severity and predict potential accident risk. The workflow included data cleaning, feature engineering, bias and fairness assessment, and model evaluation using metrics such as accuracy, F1-score, and RMSE. A user-friendly web platform was developed to provide real-time safety updates and severity predictions for citizens, with system design guided by the EU AI Act requirements for transparency, data governance, human oversight, and robustness.