This study introduces a machine learning-driven framework aimed at predicting the energy efficiency of electric city buses, with the goal of enhancing operational performance within public transit systems. At the core of this approach is a custom-built dataset that closely reflects real-world operating conditions. It encompasses features such as passenger volume, ambient temperature, HVAC usage, auxiliary power consumption, and changes in elevation. The dataset undergoes preprocessing and standardization before being used to train various regression models. Among these, the Random Forest Regressor was selected as the optimal model due to its high R² score and low Root Mean Squared Error (RMSE). The resulting predictions, expressed in kilometers per kilowatt- hour (km/kWh), provide valuable insights for stakeholders in managing costs, optimizing energy usage, and planning routes. A user-friendly Flask web application integrates the trained model, enabling real-time forecasting based on user inputs. This comprehensive implementation highlights the real-world potential of machine learning in supporting smart, energy-efficient management of electric bus fleets.