Wireless Traffic Prediction in Cellular Networks: A Study of Machine Learning Techniques

Strategic Insights -

Researchers at Port Said University have published a new study on the use of machine learning techniques for predicting wireless traffic in cellular networks. The study, titled "Machine learning techniques for spatiotemporal traffic prediction in 5G cellular networks," demonstrates the potential of machine learning algorithms in optimizing network planning and management. By analyzing the telecom Italia big data challenge dataset, the researchers developed eight models for cellular network traffic prediction, including seasonal-AutoRegressive-integrated-moving-average (SARIMA), Facebook-prophet, and Long-short-term memory (LSTM). The results show that the ensemble CNN+LSTM model is the most accurate, achieving R2 values of 0.990 for Internet, 0.986 for call, and 0.976 for SMS.

Key Takeaways:

  • The study highlights the importance of accurate and efficient techniques for predicting wireless traffic in cellular networks, enabling real-time decisions and short- and long-term forecasting.
  • The researchers developed eight machine learning models for cellular network traffic prediction, including SARIMA, Facebook-prophet, and LSTM.
  • The ensemble CNN+LSTM model achieved the highest accuracy, outperforming prior research and demonstrating enhanced predictive reliability across all network traffic types.
  • The study demonstrated the potential of machine learning algorithms in optimizing network planning and management, improving resource allocation, and ensuring high-quality service through pattern recognition in network traffic.
  • The research concluded that the ensemble CNN+LSTM model is the most accurate, followed by the hybrid CNN-LSTM and LSTM models, with the AdaBoost and XGBoost models providing practical alternatives for balancing accuracy with computational efficiency.
  • The study was conducted by researchers Alaa A. Hussien, Heba Nashaat, and Rehab F. Abdel-Kader at Port Said University.

Statistics:

  • The R2 values achieved by the ensemble CNN+LSTM model were:

+ 0.990 for Internet traffic

+ 0.986 for call traffic

+ 0.976 for SMS traffic

  • The computational time associated with the ensemble CNN+LSTM model was not specified in the study.
  • The researchers applied their models to predict different types of network traffic, including Internet, SMS, and call traffic, across distinct geographic regions: city center, commercial, residential, and business.
  • Each region exhibited unique temporal traffic patterns influenced by weekdays, weekends, and local activities.

Sources:

  • Journal of Engineering, October 20, 2025, p 2401
  • "Machine learning techniques for spatiotemporal traffic prediction in 5G cellular networks." Discover Applied Sciences, 2025,7(10):1-35.
  • Springer (publisher)
  • https://doi-org.sdpl.idm.oclc.org/10.1007/s42452-025-06746-3