Finalist
3 weeks · SNCF · Bachelor 2
Railway Incident Prediction via ML
Analyzing the correlation between weather conditions and railway incidents to predict daily defects using machine learning.
Business Context
SNCF, France's national railway company, faces weather-related infrastructure incidents that disrupt operations. Predicting these defects would enable preventive maintenance.
Strategic Problem
How to analyze the correlation between weather conditions and the frequency of railway incidents, then use machine learning to predict daily defects?
Data Sources
Historical weather data (temperature, precipitation, wind), railway incident logs, track maintenance records, and geographic infrastructure mapping.
Methodology
Analyzed seasonality patterns and weather-incident correlations. Built a predictive maintenance algorithm using the provided data for better resource allocation and maintenance personnel scheduling. Conducted time-of-day analysis on incident frequency. Proposed a climate-control algorithm that adjusts train temperature based on passenger count to reduce average climate energy consumption.
Key Results
Delivered a predictive maintenance model plus resource optimization recommendations and an energy-saving climate algorithm. Reached finalist stage.
Business Impact
Framework applicable to any infrastructure operator seeking to reduce weather-related disruptions through predictive analytics and optimize operational resource allocation.