AI-Driven Root Cause Analysis for Fixed Telecommunications Networks

Abstract

Root Cause Analysis (RCA) is a critical component of service assurance in fixed telecommunications networks, where service degradation may originate from multiple interconnected domains, including fixed access infrastructure, Optical Line Terminals (OLTs), Optical Network Terminals (ONTs), IP transport, uplinks, network capacity, Customer Premises Equipment (CPE), and home Wi-Fi environments. Traditional RCA approaches depend largely on predefined thresholds, alarm rules, individual monitoring systems, and manual correlation by network engineers. Although these methods remain effective for known and clearly defined faults, they become less efficient when network problems involve large volumes of operational data, intermittent degradation, multiple simultaneous symptoms, or dependencies across different network domains. This paper examines the application of Artificial Intelligence (AI) to enhance RCA in fixed telecommunications networks. It proposes a practical AI-driven RCA framework that combines network and service data, KPI and alarm monitoring, anomaly detection, multi-source data correlation, AI-based analysis, probable root-cause identification, corrective-action recommendations, and post-action per-formance validation. The proposed approach aims to transform RCA from a predominantly reactive and manually intensive pro-cess into a more intelligent and proactive service-assurance capability. Rather than replacing network engineers, AI-driven RCA is positioned as a decision-support mechanism that accelerates investigation, improves correlation across heterogeneous data sources, and enables more consistent identification of probable causes. The paper concludes that AI-driven RCA represents an important step toward increasingly intelligent and automated fixed telecommunications network operations.

Introduction

The text presents an AI-driven Root Cause Analysis (RCA) framework for fixed telecommunications networks. Modern fixed networks contain many interconnected components such as fiber infrastructure, OLTs, ONTs, IP transport networks, uplinks, CPEs, and home Wi-Fi. Because a single customer problem can originate from different network layers, identifying the actual cause of service degradation can be difficult.

Traditional RCA

Traditional RCA mainly depends on:

• KPI thresholds
• Network alarms
• Predefined correlation rules
• Troubleshooting procedures
• Network engineer experience

These methods work well for known and simple faults but become less effective when large amounts of data are distributed across different systems or when multiple faults occur simultaneously. Manual investigation can be time-consuming and is generally reactive.

Role of AI

AI can improve RCA by analyzing large amounts of network information and identifying relationships that may be difficult to detect manually. It can:

• Detect unusual network behavior and emerging problems.
• Correlate alarms, KPIs, topology, utilization, and customer data.
• Identify recurring patterns from historical incidents.
• Rank the most probable root causes.
• Recommend suitable corrective actions.
• Validate whether the problem was resolved.

The study emphasizes that AI should support rather than replace network engineers. Engineers should validate AI-generated conclusions before taking actions that could affect live services.

Proposed AI-Driven RCA Framework

The proposed framework contains eight major stages:

Network & Service Data → KPI/Alarm Monitoring → Anomaly Detection → Multi-Source Correlation → AI Analysis → Probable Root Cause → Corrective Action → Performance Validation

1. Network and Service Data: Collect information from OLTs, ONTs, transport networks, uplinks, CPEs, Wi-Fi systems, and customer-experience platforms.
2. KPI and Alarm Monitoring: Continuously monitor throughput, latency, packet loss, utilization, availability, optical parameters, and alarms.
3. Anomaly Detection: Use AI to identify behavior that differs from normal patterns, even when traditional thresholds have not been exceeded.
4. Multi-Source Correlation: Combine information based on time, topology, location, shared resources, and affected customers.
5. AI-Driven Analysis: Use machine learning, classification, clustering, pattern recognition, and historical incident matching.
6. Probable Root Cause Identification: Produce ranked possible causes with supporting evidence.
7. Corrective Action: Recommend actions such as capacity expansion, traffic redistribution, configuration correction, optical inspection, CPE troubleshooting, or Wi-Fi optimization.
8. Performance Validation: Check network and customer KPIs after corrective action to confirm that the problem has been resolved.

Example

If several customers connected to the same OLT experience low speeds while uplink utilization, latency, and packet loss increase, AI can correlate these conditions and rank uplink congestion as the most probable root cause rather than treating each customer complaint as an individual fault.

Conclusion

Root Cause Analysis is a fundamental component of fixed telecommunications service assurance, yet increasing network complexity and the growing volume of operational information make traditional manual and rule-based approaches increasingly difficult to scale. A single customer-visible symptom may originate from the FTTH access network, OLT/ONT infrastructure, IP transport, uplink capacity, CPE, or home Wi-Fi environment. Effective RCA therefore requires correlation across multiple network and service domains rather than analysis of individual alarms or KPIs in isolation. This paper proposed an AI-driven RCA framework that integrates network and service data, KPI and alarm monitoring, anomaly detection, multi-source data correlation, AI-driven analysis, probable root-cause identification, corrective-action recommendations, and performance validation. Such an approach can complement conventional thresholds and engineering procedures by identifying abnormal patterns, correlating heterogeneous operational information, and prioritizing probable causes for further investigation. These capabilities are consistent with the broader evolution of AI-enhanced telecommunications operation and service assurance [1], [8]. AI-driven RCA should not, however, be considered a replacement for network engineering expertise. Data quality, model accuracy, explainability, changing network conditions, and incorrect correlations remain important limitations. A practical implementation should therefore maintain engineering validation while progressively automating suitable analytical and low-risk operational activities. Overall, AI-driven RCA represents an important step toward faster, more consistent, proactive, and increasingly intelligent fixed telecommunications network operations. Its greatest value lies in converting fragmented operational data into actionable evidence that enables engineers to identify and address service degradation more effectively.

References

[1] ITU-T, Framework of Artificial Intelligence Enhanced Telecommunication Operation and Management (AITOM), Recommendation ITU-T M.3080, International Telecommunication Union, Geneva, Switzerland, Mar. 2026. [2] TM Forum, AI for Observability & Service Assurance in Autonomous Networks, IG1343, Version 2.3.0, TM Forum, 2026. [3] ITU-T, TMN Management Functions, Recommendation ITU-T M.3400, International Telecommunication Union, Geneva, Switzerland, Feb. 2000. [4] M. Mustafa et al., “Optimizing Fixed Network Performance Through AI-Driven Operations and Analytics,” International Journal for Research in Applied Science & Engineering Technology (IJRASET), 2026, doi: 10.22214/IJRASET.2026.83286. [5] M. Mustafa et al., “AI-Driven Customer Experience Optimization in Fixed Broadband Networks,” International Journal for Research in Applied Science & Engineering Technology (IJRASET), 2026, doi: 10.22214/IJRASET.2026.83484. [6] M. Liu, X. Qi, L. Liu, and H. Pan, “Roots-tracing of communication network alarm: A real-time processing framework,” Computer Networks, vol. 192, Art. no. 108037, 2021, doi: 10.1016/j.comnet.2021.108037. [7] M. Li, M. Yang, and P. Chen, “Alarm reduction and root cause inference based on association mining in communication network,” Frontiers in Computer Science, vol. 5, Art. no. 1211739, 2023, doi: 10.3389/fcomp.2023.1211739. [8] ITU-T, Requirements for Smart Comprehensive Analysis within Artificial Intelligence Enhanced Telecom Operation and Management (AITOM), Recommendation ITU-T M.3390, International Telecommunication Union, Geneva, Switzerland, Mar. 2025. [9] M. Mustafa et al., “Agentic AI for Service Assurance in Fixed Broadband Networks: Conceptual Framework for Intelligent NOC,” International Journal for Research in Applied Science & Engineering Technology (IJRASET), 2026, doi: 10.22214/IJRASET.2026.83601. [10] M. Mustafa et al., “From AI Analytics to Agentic AI: Framework for Autonomous Fixed Broadband Operations,” International Journal for Research in Applied Science & Engineering Technology (IJRASET), 2026, doi: 10.22214/IJRASET.2026.83844.

Copyright

Copyright © 2026 Mohammad Mustafa, Mohammed Hussain Moheet, Mohammed Abdul Moheet. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.