The rapid growth of mobile network traffic and the emergence of future generation mobile networks pose significant challenges in accurately predicting and managing network traffic. In this paper, we provide a unique deep-learning algorithm based approach for traffic prediction in the mobile networks. The paper proposes a novel method for predicting download bitrate in network traffic for the next-generation networks using deep learning model. Our proposed model comprises of Bidirectional LSTM coupled with an Attention mechanism. By using this mechanisms, our model makes use of the temporal and spatial trends present in the data to greatly improve prediction accuracy and flexibility to variable network conditions. MAE (Mean Absolute Error), RMSE (Root Mean Square Error), and R2 score (Coefficient of Determination) are performance metrics used to evaluate the accuracy and goodness-of-fit of regression models, with MAE measuring the average magnitude of errors, RMSE quantifying the root mean square of errors, and R2 score showing the proportion of the dependent variable’s variance that can be anticipated from the independent variables. Our suggested model outperforms the conventional ones, with an MAE of 0.00227, RMSE of 0.00469 and an R2 score value of 0.77533. Overall, the suggested model’s performance indicates that it has the potential to be used in download bitrate prediction in the real world.