Vulnerable road users (VRUs) are threatened by a proportionally higher safety risk. In order to protect VRUs more efficiently, risk models at a high spatial and temporal resolution are required. So far, potentials and limitations of the combined use of wearable sensors and urban data ecosystems for training a predictive machine-learning model for safety risks in urban environments have not been sufficiently investigated. Insights in this context could be used in a wide variety of ICT-supported application scenarios, such as transportation management, public safety planning, crowd management, health monitoring, digital mobility services or planning.