In this study, after fabricating a solar parabolic water heater, an efficient model is suggested to predict the efficiency of the solar water heater system (SWHS). Artificial neural networks (ANN) can create logical relations among the input parameters and target(s). As efficiency is trained a function of the input parameters, when conditions are desirable to measure the data, a network-trained function can be used to predict the efficiency of the solar system. The used data for the neural network analysis were measured by using experiments on a parabolic trough collector, during four days in June. Variables such as solar radiation, ambient temperature and the output fluid temperature of the collector were considered as input parameters and the efficiency of the solar parabolic water heater was used as the output neural network. Different ANN models are presented based on the various input parameters and neurons. The ANN6 model with a 4-10-1 structure, with a root mean square error (RMSE) of 0.0061 and regression coefficient for train data (Rtrain) of 0.99995, is the most accurate among the presented models. By increasing the input parameters, the RMSE decreases and accuracy of the models increases. When experimental tests are not impossible in similar conditions, the presented model can help researchers predict the efficiency of studied SWHS by saving time and costs.