The study of fuel utilization prediction explores the accuracy of machine learning algorithms in forecasting fuel utilization for bulky vehicles. The most important aspects for predicting fuel consumption are related to road grade, vehicle speed, traffic, weather condition, and so on. Machine learning methods are most successful, in forecasting fuel utilization and identifying which aspects are most dominant for fuel consumption. The main motive of this project is to increase the accuracy of the fuel utilization forecasting model with machine learning to reduce fuel utilization. By reducing the utilization of fuel there are many benefits in satisfying domain needs and business economic improvements. The new model encapsulates procedure based on distance traveled rather than the conventional methods where each individualized machine learning model is developed for fuel usage. The new model can easily be evolved for an independent vehicle in an agile in order to evaluate fuel utilization over the entire agile. This procedure is used in concurrence with seven forecasts obtained from machine speed and road grade to obtain a neural network model for fuel utilization in heavy vehicles. Forecasting of fuel utilization using a machine learning model algorithm ANN would provide better accuracy when compared to other algorithms. The forecasts of the model are accumulated over stable window sizes of mileage traveled. Dissimilar window dimensions are estimated and the outcome shows that a 1km window can forecast fuel utilization with a 0.95 quantity of persistence.