The generative component employs a tailored Generative Adversarial Network-based architecture to address the complexity of multivariate time-series generation. Such model takes as input a combination of real data and user-defined hypothetical conditions.

The What-if scenarios tab enables users to simulate scenarios based on real-world data using an this advanced artificial intelligence methodology. This tool was developed to support strategic decision-making in urban contexts, allowing users to evaluate the potential impact of significant changes in three key areas: Temporary Roads or Zones Closures, Construction of a Multistory Parking Facility, Consecutive Rainy Days. While real data is used as a foundation, the simulated scenarios do not have a direct real-world counterpart but rather represent projections based on the selected conditions.

Once each simulation runs, the relative tab displays various plots that help evaluate its impact on parking demand. These include:

• Real and Generated Data Heatmaps: displays parking activity in Caserta based on actual data and simulates parking activity under the selected scenario.
• Radar Plot: highlights the effect of the simulation on all zones, showing the redistribution of demand across the city.
• Bar Plot with Associated t-SNE: visualizes changes in parking activity and clusters zones based on their response to the simulation.
• Line Plot: compares time series trends of parking demand between generated and real data.

Using a fully data-based methodology, it is possible to explore future scenarios without real-world changes.