A number of analytical models have been described in the literature to estimate the fuel consumption of vehicles, most of which require a wide range of vehicle and trip related parameters as input data, which might limit the practical applicability of these models if such data were not readily available. To overcome this drawback, this study describes the development of three data mining models to estimate fuel consumption of a vehicle, including linear regression, artificial neural network and support vector machines. The paper presents comparison results with five instantaneous fuel consumption models from the literature using real data collected from three passenger vehicles on three routes. The results indicate that while the prediction accuracy of the instantaneous fuel consumption models varies across the data sets, those obtained by the regression models are significantly better and more robust against changes in input data.