In this work, we investigate a novel framework for a traffic-aware multi-beam optimal resource allocation to serve fixed home users distributed over a geographical area in the presence of rain and foliage. The fixed home users in the geographical area of interest are served by small-cell next-generation node B (gNB) via multiple beams generated simultaneously by the uniform planar array (UPA) installed at the gNB. In this regard, we present a framework to compute the optimal coverage radius of the gNB to satisfy the desired quality of service (QoS) requirements of the users. We also derive the closed-form expression for the optimum coverage radius of the gNB, considering free space and foliage attenuation scenarios. We propose a graphical methodology to compute the optimal radius of gNB in free space propagation, rain, and foliage attenuation. Further, based on the location information of the users, we determine the optimal location of the gNB by leveraging the unsupervised machine learning (ML) framework. Finally, we investigate a non-linear programming (NLP)-based technique for allocating optimal power and bandwidth to each beam, constrained by total power and bandwidth availability at the gNB. The optimal beam resource allocation (power and bandwidth) strategy ensures that the requested data rate (traffic demand) is satisfied for each user served by each beam. The simulation results demonstrate the effectiveness of our proposed methodology to ensure high QoS for a larger number of users in the presence of rain and foliage as compared to genetic algorithm and surrogate optimization