This paper proposes a data-driven approach to enhancing traffic safety through the integration of digital twins, in-vehicle monitoring system and machine learning. The main goal is to contribute to solving problems related to driver behavior, inadequate road signage infrastructure, and delayed maintenance by developing a digital twin model that leverages real-time data for predictive analysis, coaching, and maintenance. Using the Prophet algorithm, the model predicts compliance with traffic regulations, identifies frequent driver violations, and highlights deficiencies in road signage, enabling timely interventions. The innovation of this solution lies in its ability to synchronize real-time data from drivers, vehicles and road infrastructure and provide predictive insights, creating a scalable and adaptable framework for traffic management. The proposed model is tested in a proof-of-concept scenario, where it demonstrated significant improvements in road safety.