The development of new foaming agents without chlorine atoms has been vigorously pursued. Hydrofluoroether foaming agents have been proposed as alternatives to chlorofluorocarbons and hydrochlorofluorocarbons by the Research Institute of Innovative Technology for the Earth. Important physicochemical properties for selecting foaming agents include normal boiling points, enthalpies of vaporization, surface tensions, and thermal conductivities. This paper reports predictions of the enthalpy of vaporization using a three-layer neural network with an error back-propagation learning algorithm. Using the proposed method combined with a previously proposed predictive neural network method for boiling points and surface tensions, the molecular design of foaming agents was carried out for trichlorofluoromethane alternatives. Finally, some hydrofluoroethers were selected using thermal conductivity calculated by a multiple regression.