Energy efficiency has become a central priority in the automotive industry due to increasing environmental, economic, and regulatory pressures. This dissertation proposes and validates a predictive eco-driving assistance system (EDAS) for battery electric vehicles, designed to assist drivers in adopting a more energy-efficient driving style, specifically, while simultaneously enhancing driving safety and comfort. The proposed predictive EDAS employs eco-driving model predictive controllers that leverage data from onboard sensors, traffic signal infrastructure, and route-related geographical information to compute energy-optimal speed trajectories. These recommendations are communicated to drivers through an intuitive eco-driving feedback system providing both visual and auditory cues. To assess the effectiveness of the proposed system across diverse traffic scenarios, user studies were conducted using a dynamic driving simulator. Objective evaluations demonstrated that drivers using the predictive EDAS achieved significant energy savings, reduced speed limit violations, maintained safer speeds on curved roads, and minimized unnecessary stops at signalized intersections. Furthermore, recurrent neural network-based prediction models were developed to accurately forecast leader vehicle behavior by incorporating critical environmental information. In addition, a learning-based driver behavior modeling framework is introduced to enhance advisory speed-tracking performance and overall system efficiency. Finally, a subjective evaluation based on established acceptance models revealed an overall positive user attitude, with perceived usefulness and perceived behavioral control emerging as key factors influencing drivers' intentions to adopt predictive EDAS in daily life.