A Neuro-Fuzzy Approach for Vehicle Fuel Consumption Prediction

Abstract

This paper presents a neuro-fuzzy approach for predicting vehicle fuel consumption. The prediction of fuel consumption of a vehicle has become a strategic issue. This is because it is not only related to the problem of the availability of fuel which is getting thinner but also the problem of the environmental impact caused. In this study, the acquisition of the car parameter data was inputted, namely the number of cylinders, displacement, horsepower, weight, acceleration, and model year. The output variable that will be predicted is fuel consumption in miles per gallon (MPG). 'Weight' and 'Year' are chosen as the two best input variables. Training results and predictions are expressed in the three-dimensional input-output surface graph of the best two-input ANFIS model for MPG prediction. The graph shows a nonlinear and monotonic surface, where MPG is predicted to increase with an increase in 'Weight' and a decrease in 'Year'. The results of the RMSE training were 2.767 and the RMSE examination was 2.996. Based on the results of the study showed that the greater the weight of motor vehicles, the greater the amount of fuel needed to travel the same distance.