We developed a machine learning model for the output power of an air-cooled regenerative Organic Rankine Cycle (ORC) unit. The training data was obtained from a demonstration test conducted at the Obama hot spring in Unzen city, Nagasaki. This model was designed for prediction based on the ambient temperature and hot water temperature, while considering the mass flow rate, turbine outlet pressure, and outlet temperature. The machine-learning results indicated that the output power of the ORC unit was affected by the mass flow rate of the working fluid. Furthermore, the machine learning model demonstrated that it is possible to assess the output power using the actual ambient temperature data obtained at the Obama hot spring.