Telecom Alarm Analysis With Association Rules for Root-Cause Detection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing alarm analysis methods in telecommunications networks face inefficiencies and inaccuracies in associating alarms and locating root faults, leading to high maintenance costs and difficulty in adapting to network changes.

Innovation Solution

An alarm analysis method utilizing association rules, random forest algorithms, and sliding window modeling to efficiently analyze alarm associations and accurately identify root causes by mining key features and predicting root alarms based on subsystem and object dependencies.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If manual alarm analysis is used, then flexibility in handling complex alarm scenarios is maintained, but labor consumption increases and efficiency decreases

Engineering Contradiction:
Improvealarm analysis efficiencyVSAvoidautomation level
Core Design Contradiction:
ProductivityVSExtent of automation

Solution Approach 1:

The system performs self-service by automatically generating alarm association rules through mining frequent alarm sequences and extracting key features from alarm data. The random forest algorithm automatically predicts root causes without requiring manual rule configuration, enabling the system to autonomously handle complex alarm scenarios and reduce labor consumption while maintaining high analysis efficiency.

Inventive Principle:
Principle #25Self-service

2Reliability

If existing alarm association rules are used, then single-device level association can be achieved, but system-level association analysis is insufficient and maintenance costs are high

Engineering Contradiction:
Improvealarm association accuracyVSAvoidrule maintenance complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The system implements feedback mechanisms by continuously mining frequent alarm sequences from actual network operations and automatically updating the association rules. The random forest algorithm learns from historical alarm data and adjusts predictions based on feedback from actual fault scenarios, enabling automatic rule optimization without manual maintenance while improving both single-device and system-level association accuracy.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The system changes parameters by transitioning from static manual rules to dynamic data-driven models. The random forest algorithm processes alarm data with varying characteristics and automatically adjusts decision boundaries and feature weights, enabling the system to adapt to changing network conditions and maintain high reliability without complex rule maintenance.

Inventive Principle:
Principle #35Parameter changes

3Productivity

If alarm data is processed without feature mining, then complete alarm information is preserved, but data volume increases and processing efficiency decreases

Engineering Contradiction:
Improvedata processing efficiencyVSAvoidalarm data volume
Core Design Contradiction:
ProductivityVSQuantity of substance

Solution Approach 1:

The system extracts only the essential information by mining frequent alarm sequences and identifying key features that are most predictive of root causes. The random forest algorithm selectively processes only the most relevant alarm attributes, filtering out redundant information while preserving critical diagnostic data, thereby reducing data volume and improving processing efficiency without sacrificing analysis accuracy.

Inventive Principle:
Principle #2Taking out (Extraction)

Data Source

PatentEP3965371B1Alarm analysis method and related device
Publication Date: 2025.07.09 HUAWEI TECH CO LTD
  • EP3965371B1 patent drawingFigure 1
  • EP3965371B1 patent drawingFigure 2
  • EP3965371B1 patent drawingFigure 3

AI summary

Embodiments of the present invention disclose an alarm analysis method and a related device. The method includes: determining M alarm pairs in a first alarm set, where each of the M alarm pairs includes a first alarm and a second alarm between which an association exists; generating, according to an association rule, a first feature set of N alarm pairs, in which the first alarm is used as an alarm pair root, in the M alarm pairs, where the first feature set includes a first probability that a first subsystem to which each first alarm in the N alarm pairs belongs is a subsystem root and a first alarm object corresponding to the first alarm is an alarm object root and a second probability that a second subsystem to which each second alarm belongs is a subsystem root and a second alarm object corresponding to the second alarm is an alarm object root; and determining root information of the first alarm set based on the first probability and the second probability. According to this application, an association between alarms can be efficiently analyzed and a root alarm can be accurately located.