A comparative study of the performances of three prediction models (BiLSTM, Holt-Winters and SARIMA) was carried out to estimate the traffic of a mobile EDGE network. Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) are used to evaluate the models. Our results on voice and data traffics show that the BiLSTM model outperforms the other two models studied in estimating EDGE network traffic, indicating its potential as a useful tool for telecommunications service providers. For voice traffic, we obtained an average MAE about 0.73 and an average RMSE about 1.0. This study contributes to continuing efforts to improve network performance, enable efficient use of available resources and ensure a high quality of service in the ever-changing telecom sector.