Home > Research > Publications & Outputs > AI-Driven Data Analytics and Intent-Based Netwo...

Associated organisational unit

Electronic data


Text available via DOI:

View graph of relations

AI-Driven Data Analytics and Intent-Based Networking for Orchestration and Control of B5G Consumer Electronics Services

Research output: Contribution to Journal/MagazineJournal articlepeer-review

E-pub ahead of print
  • Khizar Abbas
  • Ali Nauman
  • Muhammad Bilal
  • Jae-Hyung Yoo
  • James Won-Ki Hong
  • Wang-Cheol Song
<mark>Journal publication date</mark>13/10/2023
<mark>Journal</mark>IEEE Transactions on Consumer Electronics
Publication StatusE-pub ahead of print
Early online date13/10/23
<mark>Original language</mark>English


Network slicing is a critical feature of the beyond fifth-generation (B5G) network that supports a wide range of innovative services from 5.0 industries, next-generation consumer electronics, smart healthcare, etc. Network slicing guarantees the provisioning of quality of service (QoS) aware dedicated resources to each service. However, the orchestration and management of network slicing is very challenging because of the complex configuration process for underlying network resources. Furthermore, the third generation partnership project (3GPP) presented artificial intelligence (AI) based network data analytics function (NWDAF) in 5G for proactive management and intelligence. Therefore, we have developed an intent-based networking (IBN) system for automating network slices and an AI-driven NWDAF for proactive and intelligent resource assurance. The network data analytics function uses a hybrid stacking ensemble learning (STEL) algorithm to predict network resource utilization and a novel automated machine learning (AutoML) and voting ensemble learning-based mechanism to detect and mitigate network anomalies. To validate the performance of the implemented work, real-time datasets were employed, and a comparative analysis was conducted. The experimental result shows that our STEL model enhances the accuracy by 20% and reduces the error rate by 45%. The AutoML and ensemble learning-based optimized model achieved 99.22% accuracy for anomaly detection.