This paper aims to present a model for traffic risk prediction. Its contribution is the adaptation of Fuzzy Logic applied to Harmonic Systems in order to make it more flexible and powerful in certain contexts. The possibility of having a good traffic risk prediction opens a practical possibility to successfully improve the security not only for drivers, but also for pedestrians and cyclists. The proposed model is able to process in real-time with simple data provided by the environment and the individual whose risk is being processed. The scope of this paper covers the technical description of the model, statistical analysis and comparison with alternates using Traditional Ruled-Expert Systems (RES), Harmonic Systems (HS) and Fuzzy Harmonic Systems (FHS). Also a short proposal for the prototype is described. Results indicate a remarkable improvement for the FHS predictor compared to RES and HS.