Intelligent Traffic Control System is an integral part of modern transportation system, helping to maintain smooth and safe traffic flow while reducing pollution. In this paper, we propose an intelligent traffic control system for multi-intersection networks that aims to improve traffic control systems by adjusting signal light timings to reduce waiting times and considering vehicle types and priorities. The model combines RL to obtain optimal control policies, GCN to capture spatial dependencies, LSTM to capture temporal dependencies, and GA to enhance the deep network weights quickly and escape local optima. The experiment evaluates the effectiveness of various RL-based models in traffic management by evaluating the impact of GA and prioritization on ITCS models. Models are trained/tested using synthetic traffic data generated with the SUMO tool on three different-sized networks: Manhattan, Suzhou, and Cairo, with various vehicle types. The results demonstrate the distinct improvements of the LSTM-GCN-GA model in reducing waiting times. When compared with traditional models such as the Pre-Time model as in the Manhattan network, it reduced the waiting time by up to 84.81% for all vehicles and by up to 92.46% for priority vehicles. The GA integration reduced the waiting time by up to 26.39% for all vehicles and by up to 80.21% for priority vehicles. Adding vehicle priority reduced waiting time by up to 33.1% for all vehicles and by up to 83.82% for priority vehicles. Applying this model in real-world applications can enhance neural network efficiency, which optimize traffic flow, reduce congestion, and improve road safety.