Path loss prediction is a key component in constructing wireless networks allowing for optimal base station placement and configuration. In this paper, we propose a DNN model to predict path loss with higher precision than that of current models (e.g., COST-231) in urban environments. The outcomes indicate that the suggested model reduces error rates significantly and high precision compared to traditional methods by a wide margin, making it more robust under a variety of circumstances. Residual analysis corroborates the stability of the predictions, with errors centered at the zero mean. Moreover, the adaptability of the model makes it ideal for diverse applications in heterogeneous settings. This work is a summary of the great success and remarkable advancement of wireless communication with deep learning, resulting from the static format of traditional equations and maximization of the function to a variable function, thus enhancing the performance of the system or the network, which ultimately leads to the request operation of low price and high quality.