The final goal of this study is to develop a “data-driven cavitation model” based on a machine learning model that is trained by the measurement data, rather than a conventional mathematical model, aiming at improvement of the model by the data assimilation aspect of using measured data. In this paper, as a preliminary study, a framework for a data-driven cavitation model is developed by employing CFD data calculated with a mathematical homogeneous fluid model as the training dataset. The target is cavitating flow around a Clark-Y11.7% hydrofoil at an angle of attack of 0 degree. The neural network for data-driven cavitation model is U-Net, which is a kind of convolutional neural network. The training dataset consists of the fields of the velocity, pressure, and liquid volumetric fraction f_L of the current timestep and those of the next timestep. In the test mode, trained U-Net is generally able to predict the f_L field of the next step except at the trailing edge vortex where the cavity rapidly grows. The data-driven cavitation model implemented in the flow simulation code quantitatively reproduces the same flow field as that of the mathematical model. However, computation becomes unstable where cavity grows rapidly and pressure waves are generated due to cavity volume fluctuations, suggesting that a data-driven cavitation model needs further improvements to robustly capture such highly unsteady phenomena.