Truck Fuel Consumption Prediction Using Logistic Regression and Artificial Neural Networks

Abstract

Rising international oil costs and the transport industry's recovery from the effects of Covid-19 resulted in the efficient management of fuel by logistics companies becoming a significant concern. One way of managing this is by analyzing the fuel consumption of trucks so as to better utilize the costly resource. Twenty-three driving data variables were gathered from 210 freight trucks and analyzed this data. Relevant variables that impact truck fuel consumption were extracted from the initial 23 variables gathered using stepwise regression, and then a prediction model was built from the identified relevant variables utilizing a binary logistic regression model. In addition, a back propagation neural network was employed in this study to create a second model of truck fuel use, and comparisons between the two models were made. The outcomes showed that the binary logistic regression model and the back-propagated neural network model prediction accuracy were 68.4% and 77.2%, respectively.