An Enhanced Traffic Matrix Prediction in Software-Defined Networks Using Bi-LSTM with Attention Mechanism

Abstract

Software-defined networks (SDNs) play a vital role in network traffic management, traffic estimation, traffic engineering, and routing. The effective management of SDNs relies on traffic matrix prediction (TMP). The traffic matrix (TM) provides a comprehensive view of each origin–destination flow. An SDN paradigm offers various techniques to obtain additional information about the TM. This information is useful for managing network operations, such as performance diagnostics, traffic engineering, and network design. The TMs are critical for network operation and management. However, TMP remains a challenging task due to the dynamic and complex nature of network traffic patterns. To address this challenge, a Bi-LSTM model with attention mechanism is proposed for better prediction. The proposed model captures both forward and backward temporal dependencies in historical traffic data that enable the model to learn short-term and long-term traffic correlations, while the attention mechanism focuses on critical timesteps that help the model learn the traffic behavior and traffic patterns more effectively. The proposed model is experimented with real-world topologies such as Abilene, GÉANT, Nobel-Germany, and Germany50. The TMs of these topologies are in the flattened form with different granularities and horizons. The performance of the proposed model is evaluated using metrics such as mean squared error, mean absolute error, coefficient of determination, and average inference time. The experimental results show that the Bi-LSTM model with attention mechanism has outperformed the Bi-LSTM with Adam model in terms of accuracy with better TMP in all considered real-world topologies.

Received: 20 July 2025 | Revised: 9 January 2026 | Accepted: 2 February 2026

Conflicts of Interest

The authors declare that they have no conflicts of interest to this work.

Data Availability Statement

Data sharing is not applicable to this article as no new data were created or analyzed in this study.

Author Contribution Statement

Prabu. U.: Conceptualization, Methodology, Software, Validation, Investigation, Resources, Data curation, Writing – original draft, Visualization. Prabhala Harichandana: Conceptualization, Methodology, Software, Validation, Investigation, Resources, Data curation, Writing – original draft, Visualization. Manepalli Baby Kavyasri: Conceptualization, Methodology, Software, Validation, Investigation, Resources, Data curation, Writing – original draft, Visualization. Uppala Bhargavi: Conceptualization, Methodology, Software, Validation, Investigation, Resources, Data curation, Writing – original draft, Visualization. Geetha. V.: Formal analysis, Writing – review & editing, Supervision, Project administration.