Mm-wave communication using massive Multiple-Input Multiple Output (MIMO) is highly directional in nature, hence making beam scanning and alignment a crucial aspect. In high mobility scenarios, frequent beam scanning may incur a performance degradation due to high control overhead. In this paper, we have adopted a supervised learning approach to determine the precoders and combiners for a hybrid beamforming architecture. To avoid frequent beam scanning and switching, we propose an overhead-aware strategy that addresses a trade-off between the spectral efficiency (SE) and the overhead signalling. Simulations are performed using channel data generated using the open source channel simulator NYUSIM.