Among the global problems emanating from vehicular traffic across various cities, which include but not limited to, security, parking, pollution and congestion and with the fact of the impracticality of changing the infrastructure of urban area, researchers in most cases consent that such problems may be alleviated by correctly staging the traffic lights which helps in improving the flow of vehicles across cities. In order to efficiently solve this congestion issue, reduce the total journey duration experienced by each vehicle, increase the total number of vehicles reaching their destination per unit of time and to decrease the amount of pollution emitted by the moving vehicles, this problem is addressed. Two different approaches are proposed, the first approach is a proactive approach based on Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM) to predict the traffic congestion in advance by training the machine learning models with the collected historical data. The second proposed approach is a real time reactive approach which is using different meta-heuristics and compare them together to find the near optimal solution for traffic light cycle time.