Fuel cells eliminate pollution caused by burning fossil fuels; hence, a proton exchange membrane fuel cell (PEMFC) is one of the promising technological advances for the future of the transportation industry. The key existing challenges for fuel cell commercialization are performance, design and vehicle efficiency. Since the analytical model expressing fuel cells' characteristics is not accurate in comparison with the real system's performance a robust and dynamic model for fuel cells is of great importance. This study aims to introduce an optimized model for PEMFC using an electric bicycle that consists of a 250 W fuel cell, battery pack, DC/DC convertor, electric motor and electric control unit (ECU). In the first phase of this multi-fold study, the analytical model of PEMFC's efficiency has been compared with the experimental results obtained from the electric bicycle. The result of this phase showed an overall system efficiency of 35.4% and a maximum fuel cell efficiency of 63%. This confirms that fuel cell performance is least efficient when functioning under maximum output power conditions. In the second phase of this research, the collected data was used for developing linear and nonlinear regression models. The resulting model was compared with an artificial neural network used for the same purpose, and their prediction efficiencies compared. Results show that neural network modelling improves accuracy and provides promising performance for the electric bicycle system.