In highly dynamic Vehicular Ad Hoc Networks (VANETs), maintaining reliable and efficient handover decisions is critical for ensuring stable connectivity between vehicles and Roadside Units (RSUs). In this paper, we propose and implement a handover decision-making system for VANETs considering Genetic Algorithm (GA) and Reinforcement Learning (RL) called GA-RL handover decision system. In the first phase, the GA is employed to globally optimize the weighting parameters used in handover decision-making, considering key metrics such as signal strength, RSU load, time-to-stay, and penalty cost. These GA-optimized weights are then integrated into RL using a Deep Q-Network (DQN) framework, where the agent learns a dynamic handover policy by interacting with the vehicular environment. The training process utilizes a replay buffer and target network to improve convergence stability and performance. From the simulation results, the GA-RL handover decision system consistently performs better than the traditional A3-RSRP method across multiple key performance indicators. Also, it achieves higher handover success rates, lower failure rates, reduced handover frequency volatility, fewer packet drops, and significantly better throughput retention during handover. These improvements indicate that GA-RL system provides more stable, adaptive, and efficient handover decisions, making it a good approach for maintaining better communication performance and seamless mobility in VANET environments.