This work presents a novel solution to the issue of traffic forecasting in sensorless urban areas using Floating Car Data (FCD). FCD is scalable and widespread and offers the potential for accurate traffic flow estimates for areas without fixed traffic sensors through the use of machine learning algorithms. Using FCD and a limited set of fixed sensors for training, a forecasting model is developed, enabling the estimation of traffic flows on major urban roads up to four hours in advance. This method allows traffic managers to forecast conditions throughout the network at a lower cost than using physical sensors. Preliminary results from the case study of Catania (Italy) suggest that the proposed methodology has the potential to improve traffic management with limited investment costs.