This study addresses the major urban challenge of traffic congestion, causing a considerable increase in travel time, fuel consumption, air pollution, and commuter stress. With a rapidly increasing urban population in cities like the metropolitan cities of India, this problem has taken a serious turn, especially during peak hours. Traditional traffic signal systems are often static and inefficient in such situations. This paper proposes an adaptive traffic signal control system for the adjustment of traffic signal duration based on historical traffic patterns, utilizing predictive models (ARIMA, LSTM, GRU, Bi-GRU, Bi-LSTM) through the integration of simulation tools like SUMO and real-time control via TraCI. These models are evaluated based on their performance in reducing queue length, average waiting time, and overall travel time. These results demonstrate the effectiveness of adaptive traffic signals in improving urban traffic management. Bi-LSTM was found to be the best fit for the problem by 40.08% reduction in average waiting time, 43.31% reduction in queue length, and a 17.90% reduction in travel time, despite exhibiting a higher Mean Squared Error (MSE) in traffic flow prediction. A combined CO₂ reduction metric was also assessed to calculate the impact of these predictive models on environmental sustainability. Bi-LSTM achieves a CO₂ emission reduction of 33.32%, which positions it as the most effective model for traffic control.