Weather Delay 🚂⛈️

🚂⛈️ Overview

Weather Delay is a machine learning project focused on predicting train delays from historical delay behavior and weather conditions.

The broader idea is simple: if delay predictions become more reliable, rail travel becomes easier to trust, and that can support a shift toward more sustainable transportation.

Why I built it

Delays are one of the main pain points for train users. I wanted to explore whether weather signals, combined with operational delay patterns, could improve predictive quality enough to be practically useful.

The project aims to:

• •Improve delay estimates for passengers
• •Support operator-side planning decisions
• •Increase trust in rail as an alternative to car travel
• •Contribute to sustainability-oriented mobility strategies

🦾 Repository

Code and notebooks: github.com/gianlucarea/weather-delay

📊 Methodology

The study evaluates both classification and regression pipelines:

1. •Classification: Predict if a train will be delayed beyond a threshold
2. •Regression: Predict the exact delay duration

Both approaches model relationships between:

• •Departure delay (key historical indicator)
• •Weather conditions at the time of travel
• •Route and temporal factors

Modeling workflow

1. •Collect and align train and weather datasets.
2. •Clean and preprocess temporal and route-level features.
3. •Engineer predictive features from delay and weather signals.
4. •Train and compare classification and regression models.
5. •Evaluate performance tradeoffs and feature impact.

📈 Data Sources

The project combines multiple data providers:

• •Weather Data: Open-Meteo - Free weather API with historical data
• •Train Data: Tuscany Train Feed - Regional train schedules and operations
• •Delay Information: Italian Train Delay Data - Historical delay records

🛠️ Tech Stack

• •Python: Core implementation language
• •Pandas: Data manipulation and analysis
• •NumPy: Numerical computations
• •Scikit-Learn: Machine learning algorithms and evaluation
• •Jupyter Notebooks: Interactive data analysis and experimentation

🚀 Project Structure

Main analysis artifact:

• •02-Data Analytics: Main data analysis and model development notebook
  • •Data loading and exploration
  • •Feature engineering from weather and delay data
  • •Model training and evaluation
  • •Comparison of classification and regression approaches

Note: data files must be downloaded from the listed sources before preprocessing and training.

📝 Research Focus

• •Comparing different classification algorithms for delay prediction
• •Evaluating regression models for delay duration estimation
• •Feature importance analysis of weather and temporal factors
• •Model performance trade-offs between accuracy and simplicity

What I learned

• •How weather and operational signals interact in delay prediction tasks
• •How to frame the same domain problem as both classification and regression
• •How to communicate model tradeoffs clearly in notebook-based experiments

🤝 Contributing

Contributions are welcome. If you want to improve data preparation, model evaluation, or documentation clarity, feel free to open a pull request.

Suggested contribution flow:

1. •Fork the repository
2. •Create a new branch for your feature or bugfix
3. •Commit your changes and push them to your fork
4. •Submit a pull request with a detailed description of your changes

Please ensure that your code follows the project's coding standards and includes appropriate documentation.

📜 License

This project is licensed under the MIT License. See the LICENSE file for more details.