As interest in sustainable buildings increases, demand for natural ventilation (NV) systems and their efficient operation methods are increasing. For the successful implementation of natural ventilation systems in buildings, it is essential to clarify when and how to use natural ventilation systems in advance. However, it is difficult to predict accurate NV rate based on actual data due to the uncertainty of NV and the difficulty in measuring sufficient data. In order to overcome the limitation, this study utilized transfer learning (TL) method to establish a prediction model of NV rate with even insufficient real datasets. The performance of NV prediction model based on TL was showed relatively high accuracy despite the insufficient amount of data.