"Emphasizing the significance of fuel consumption across all vehicle lifecycle stages, this paper proposes a data summarization strategy centered on distance rather than conventional time intervals for crafting personalized machine learning models. The focus is on optimizing fuel consumption in heavy vehicles using machine learning techniques. The effectiveness of a methodology aimed at reducing fuel consumption in heavy-duty vehicles (HDVs) is detailed and assessed through simulations and real-world HDV experiments. The suggested model can be easily tailored and implemented for each vehicle within a fleet to enhance overall fuel efficiency. Furthermore, the study demonstrates the reliability of simulations for direct application to real HDVs. Notably, in scenarios where speed variation range was limited, the proposed method exhibited an average improvement of approximately 3 percentage points over standard predictive cruise control (PCC) across identical road profiles."