Electric vehicle (EV) driving range prediction is crucial for enhancing EV adoption and mitigating range anxiety among drivers. Despite advancements in battery technology, accurately estimating the remaining driving range under varying conditions remains a significant challenge. To address this, we propose a novel approach capable of prognosticating the Maximum Driving Range (MDR) an EV can achieve. Our method intelligently segments routes and integrates machine learning with physics-based models to predict vehicle speed, energy consumption, and power usage, offering a more precise and dynamic driving range estimation. Specifically, the approach utilizes stochastic dropout-based Long Short-Term Memory (LSTM) networks to predict vehicle speed, which serves as input to a Light Gradient Boosting Machine (LightGBM) model for estimating energy and power consumption. In addition, the Maximum Driving Range (MDR) concept is introduced as a new metric for identifying “hazard zones” where the likelihood of battery disconnection is high. Our approach was validated through real-world driving tests across three case studies in San José, Costa Rica. The model achieved a mean absolute error of 6.87 km/h for speed, 0.067 kWh for energy consumption, and 8.57 kW for power consumption, successfully predicting hazard zones 3.90 to 8.70 km before battery disconnection. These findings demonstrate the potential of this hybrid model in enhancing EV range predictions, offering a practical tool for improved route planning and driver confidence. Future work could involve adapting the model to diverse driving scenarios, incorporating user-specific risk tolerance and environmental factors, and improving computational efficiency to support real-time applications.