Environmentalists have been more worried about the effects of increasing traffic in recent decades, particularly about the release of greenhouse gases and the widespread problem of noise pollution caused by vehicles on the road. The state of our health and happiness has been profoundly affected by these difficulties. Implementation is still unclear despite massive international studies on noise mapping and possible solutions. Reducing background noise is a critical step in creating long-lasting communities. However, it will be an in-depth understanding of noise’s spatial and temporal dynamics. A noise prediction model to evaluate the Leq on a monthly and daily basis is hence the goal of this work. Using actual traffic data that includes a variety of traffic characteristics, we explore the use of deep learning for traffic noise modeling. To train the LSTM model, 2020 Delhi area noise data from the Central Pollution Control Board was randomly split into 80% training, 10% validation, and 10% testing sets. The LSTM model outperformed the suggested regression model and the existing classical approaches in a head-to-head comparison. The research confirmed that noise pollution is quite common in the region under consideration, urging for immediate action. In line with the expanding data resources of smart cities, the trained model offers real-time forecasts of traffic noise by utilizing data from traffic monitors. This reflects well for future smart city projects. Early decision-making by policymakers is made possible by proactive noise exposure forecasts, which open the door to solutions for reducing noise levels.