🎓 Bachelor Thesis · Forecasting Tourist Flows with Machine Learning

My thesis explored how machine learning models can predict tourist activity across points of interest using multi-year data from 2014 to 2022. The aim was to compare different forecasting architectures, from classical statistical methods to modern deep learning approaches, and analyze how each captures temporal dynamics, seasonality, and irregular patterns.

A key aspect of the project was preserving privacy while maintaining data realism through the creation of synthetic datasets. This led to experiments involving clustering techniques, temporal embeddings, and hybrid loss functions to assess model robustness and sensitivity.

Under the guidance of Sara Migliorini (supervisor) and Elisa Quintarelli and Alessandro Farinelli (co-supervisors), the project evolved into more than a technical benchmark. It became an exercise in precision, interpretation, and scientific discipline. The experience taught me how data can reflect the rhythm of human movement.

⸻

Period: Oct 2023 – May 2024
Supervisors: Sara Migliorini, Elisa Quintarelli, Alessandro Farinelli

Focus Areas

• Temporal forecasting
• Synthetic data privacy
• Model interpretability

Tools

• Python, Pandas, NumPy
• PyTorch, scikit-learn