A novel machine learning methodology, BOOST-RPF, has been introduced to overcome critical scalability and generalization issues plaguing power flow analysis in modern electrical distribution systems. Current classical solvers often encounter limitations when processing large-scale, intricate grids, while many existing machine learning models struggle to consistently perform across diverse network topologies. BOOST-RPF fundamentally re-engineers the problem of voltage prediction, moving away from a conventional global graph regression task. Instead, it frames voltage forecasting as a sequential path-based learning problem1. This is achieved through a meticulous decomposition of radial networks into their constituent root-to-leaf paths, enabling a more modular and robust prediction mechanism. This innovative framework promises enhanced accuracy and adaptability in power flow assessments, which are vital for maintaining grid stability, optimizing energy distribution, and facilitating the integration of renewable sources. Advancements in these analytical capabilities are paramount for ensuring the resilience and operational efficiency of critical national infrastructure, directly impacting energy security and sustainability initiatives.