Sign in Subscribe
Artificial intelligence

Hybrid Recurrent Neural Network Architecture for Subscriber Traffic Prediction

Strategic Insights

— 1 min read

Research in the field of applied computational intelligence and soft computing has led to the development of a new hybrid recurrent neural network architecture for predicting subscriber traffic in mobile telecommunication networks. The proposed architecture combines features of Jordan and Frasconi networks and leverages incremental learning through restricted Coulomb energy at the local level, while also benefiting from global learning through backpropagation. The hybrid model was tested on a dataset of cellular traffic data collected from a telecommunication operator in Cameroon, demonstrating superior performance compared to traditional Jordan and Frasconi networks.

Key Takeaways:

  • The proposed hybrid recurrent neural network architecture combines features of Jordan and Frasconi networks to improve subscriber traffic prediction in mobile telecommunication networks.
  • The architecture integrates two sublayers, a sigmoid and a Gaussian function, within a hidden layer that forms a loop, enabling incremental learning through restricted Coulomb energy at the local level.
  • The hybrid model was tested on a dataset of cellular traffic data collected from a telecommunication operator in Cameroon and outperformed traditional Jordan and Frasconi networks with a root mean square error (RMSE) of 14.18 and a mean absolute error (MAE) of 8.61 using the Adam optimizer with a batch size of 64.
  • The hybrid model demonstrated superior performance in terms of RMSE, MAE, and convergence speed compared to existing models.
  • The proposed model has the potential to support telecom operators in proactive congestion management and resource optimization.
  • The research concluded that the hybrid model could be a valuable tool for telecom operators in managing subscriber traffic and improving network efficiency.

Statistics:

  • RMSE of 14.18, indicating a moderate level of error in subscriber traffic prediction.
  • MAE of 8.61, indicating a relatively low level of error in subscriber traffic prediction.
  • Convergence speed of the hybrid model was improved compared to existing models.
  • The dataset used for testing the hybrid model consisted of cellular traffic data from a telecommunication operator in Cameroon.
  • The Adam optimizer with a batch size of 64 was used for training the hybrid model.

Sources:

  • A Hybrid Recurrent Neural Network Architecture for the Prediction of Subscriber Traffic in a Mobile Telecommunication Network. Applied Computational Intelligence and Soft Computing, 2025.
  • Department of Computer Science.
  • Applied Computational Intelligence and Soft Computing. Wiley.
  • NewsRx. Department of Computer Science Researchers Detail Findings in Applied Computational Intelligence and Soft Computing. Network Weekly News. September 15, 2025; p 32.

Read more