Graph Neural Network based Root Cause Analysis Using Multivariate Time-series KPIs for Wireless Networks

Yen, Clive H.; Sun, Wei; Purmehdi, Hakimeh; Park, Won; Deshmukh, Kunal; Thakrar, Nishank; Nassef, Omar; Jacobs, Alice K.

doi:10.1109/noms54207.2022.9789858

Cited by 6 publications

(2 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It used KPIs data and a graph structure learning (GSL) model to identify depen-dencies and potential root causes of failures Ref. [27]. This method combined KPIs with knowledge of wireless network failures to infer hidden information and explore the potential relationship between failures and root causes.…”

Section: B Knowledge and Data-driven Failure Localization Methodsmentioning

confidence: 99%

FLAM: Locating and Mitigating 5GC Network Failure with Knowledge Graphs in China Telecom’s Network

Yu,

Hou,

Yuan

et al. 2024

Preprint

View full text Add to dashboard Cite

China Telecom’s fifth-generation core (5GC) network is complicated owing to its characterizations of service-based architecture and network functions virtualization. Thus, it is vulnerable to network failures. When a failure occurs, operations and maintenance (O&M) experts need to first analyze the root cause based on their professional experience, and then recommend an available solution for the failure. However, 5GC network failures occur frequently, and most of them are similar. Thus, inviting O&M experts to the 5GC network scene is expensive and time-consuming (hourly level). In this paper, we propose a knowledge and data-driven 5GC network called failure location and automated mitigation (FLAM) mechanism. Particularly, FLAM demonstrates the expertise in dealing with various network failures by using knowledge graphs. Four state-of-the-art machine learning algorithms were compared in FLAM to determine which one can better locate the root cause of network failures. A real-time checking module was also designed to automatically diagnose the related network functions for network failures. Based on China Telecom’s real-wild data of network failures, the proposed mechanism was evaluated considering in the metrics of algorithm complexity and location accuracy. Experimental results show that the decision tree model had an accuracy of ∼ 99% for locating the root cause of network failures, outperforming the random forest, support vector machine, and k-nearest neighbor algorithms.

show abstract

Section: B Knowledge and Data-driven Failure Localization Methodsmentioning

confidence: 99%

FLAM: Locating and Mitigating 5GC Network Failure with Knowledge Graphs in China Telecom’s Network

Yu,

Hou,

Yuan

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

“…In terms of the wireless network field, Ref. [26] proposed a graph neural network (GNN) based RCA framework that used KPIs data and a graph structure learning (GSL) model to identify dependencies and potential root causes of failures. This method utilized KPIs together with wireless network failures knowledge to infer hidden information and explore the potential relationship between failures and root causes, which can successfully identify root causes of wireless network failures for further mitigation actions.…”

Section: Knowledgeanddata-driven Failure Localization Methodsmentioning

confidence: 99%

FLAM: Locating and Mitigating 5GC Network Failure with Knowledge Graphs in China Telecom’s Network

Yu,

Zhang,

Jiang

et al. 2023

Preprint

View full text Add to dashboard Cite

China Telecom’s fifth-generation core (5GC) network is complicated owing to own characterizations of service based architecture and network functions vir-tualization, and thus is vulnerable of network failures. When a failure occurs, operations and maintenance (O&M) experts need to firstly analyze its root cause based on their professional experience; then recommending an available solution for the failure. However, 5GC network failures occur frequently, and most of them are similar; thus, inviting O&M experts to the 5GC network scene costs a super-long of time (hourly level) and wastes too much money. In this paper, we propose a knowledge&data-driven 5GC network Failure Location and Automated Mitigation (FLAM) mechanism. Particularly, FLAM illustrates the experts experience of different network failures using knowledge graphs. Four state-of-the-art machine learning algorithms are applied in FLAM to compare which one can better locate the root cause of network failures. Besides, a real-time checking module is designed to automatically diagnose the related network functions for network failures. Based on China Telecom’s real-wild data of network failures, the proposed mechanism is evaluated in the metrics of algorithm complexity and location accuracy. Experimental results show that the decision tree model has an accuracy of 99% for locating the root cause of network failures, which owns the best performance compared to the random forest, support vector machine, and k-nearest neighbor algorithms.

show abstract

Comprehensive fault diagnosis for multiple coupled SFCs based on deep learning

Xia,

Liu,

Wang

et al. 2024

Computer Networks

View full text Add to dashboard Cite

Graph Neural Network based Root Cause Analysis Using Multivariate Time-series KPIs for Wireless Networks

Cited by 6 publications

References 12 publications

FLAM: Locating and Mitigating 5GC Network Failure with Knowledge Graphs in China Telecom’s Network

FLAM: Locating and Mitigating 5GC Network Failure with Knowledge Graphs in China Telecom’s Network

FLAM: Locating and Mitigating 5GC Network Failure with Knowledge Graphs in China Telecom’s Network

Comprehensive fault diagnosis for multiple coupled SFCs based on deep learning

Contact Info

Product

Resources

About