The automotive industry has to forecast vehicle fuel economy because of concerns about pollution control, environmental sustainability, and fuel consumption. Standardized driving cycles and laboratory testing are costly, time-consuming, and usually do not accurately represent real-world driving scenarios and nonlinear vehicle parameter interactions. To get beyond these restrictions, this research proposes a machine learning approach based on ensemble learning for accurate fuel efficiency prediction. Pre processed and aSnalyzed historical vehicle datasets include engine displacement, horsepower, weight, acceleration, and model year. Prediction accuracy is increased by using Random Forest and Gradient Boosting algorithms to model complex, nonlinear feature relationships. Performance is compared using criteria like accuracy and inaccuracy. According to experimental results, the Gradient Boosting model has a higher predictive power with an accuracy of 80% compared to the Random Forest model's 75%. Scalable, data-driven, and economical, the proposed fuel efficiency estimating approach supports sustainable, energy-efficient transportation systems by assisting manufacturers, legislators, and end users in making informed decisions.