Fuel Consumption Models Applied to Automobiles Using Real-time Data: A Comparison of Statistical Models

Abstract

AbstractEven though the number and variety of fuel consumption models projected in the literature are common, studies on their validation using real-life data is not only limited but also does not fit well with the real-time data. In this paper, three statistical models namely Support Vector Machine (SVM), Artificial Neural Network and Multiple Linear Regression are used in term of prediction of total and instant fuel consumption. The models are compared against data collected in real-time from three different passenger vehicles on three routes by causal drive, using a mobile phone application. Our outcomes reveal that, the results obtained by the models vary depending on the total consumption and instant consumption correlation. Support Vector Machine model of fuel consumption expose comparatively better correlation than the other statistical fuel consumption models.