Road safety analysis is vital for ensuring adequate protection across networks that accommodate both vehicles and active road users, including bicycle ways and pedestrian paths at bus stops. Despite the growing availability of data describing crash events and infrastructure characteristics, there is limited empirical evidence on how those characteristics influence the occurrence of crashes involving cyclists and pedestrians. The aim of this thesis is to address this gap by developing Generalized Linear Models for crash prediction using historical crash data and infrastructure geometric and functional data for the Metropolitan City of Milan. In the case of bicycle ways, predictive models were developed for intersections with roadways and segments, while for bus stops, the analysis was focused on their influencing area.