A Practical Framework for Generating and Validating Synthetic Databases: Application to Freight Transport

Synthetic databases are increasingly used in research and industry to support testing, training, and analysis without exposing sensitive information. This paper proposes a practical framework for generating and validating synthetic databases, structured around a pipeline that ensures structural consistency, business relevance, and reproducibility. The framework is illustrated through a case study on freight transport in Romania, where a relational model was designed to capture entities such as clients, trains, conductors, and transported goods. A Python-based generator was developed to populate the database with realistic values under domain-specific constraints (e.g., valid national identifiers, capacity limits, distinct departure/arrival stations). Validation is focused on structural integrity, query performance, and privacy preservation. The results show that the generated dataset is both realistic and safe for academic or enterprise use, while the methodology is transferable to other economic and business contexts.