Fifth-generation mobile technology provides ultra-reliable low-latency communication (URLLC) for mission-critical applications, massive machine-type communication (mMTC) for IoT connectivity, and enhanced mobile broadband (eMBB) services for high-speed data transfer. To accomplish this goal, beamforming plays a crucial role in radio resource allocation by directing a focused beam toward the user. Some of the beamforming mechanisms calculate the optimal beam pair index based on the channel state information (CSI) and signal to interference and noise ratio (SINR). To identify the optimal beam pair index in the beam selection procedure, it is essential to employ a unique technique that minimizes overhead during beam sweeping and selection while maintaining low complexity. In this paper, we used neural networks for beam selection based on the Global Positioning System (GPS) coordinates of the receiver. Neural networks take GPS coordinate of the receiver and optimal beam pair index as input to train the model. In the output of the neural network, the K optimal beam pair indices are chosen based on the average reference signal received power (RSRP). Neural networks contribute to the high accuracy and average RSRP beam selection as compared to the benchmarked algorithm.