Cloud Network Alarm Root Cause Relational Tree Model Construction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional methods for determining alarm root-cause relations in communication networks, especially in 5G networks, are inadequate due to the complexity introduced by virtualization and slicing technologies, leading to difficulties in updating knowledge bases and effectively modeling alarm root-cause relational trees for new networks and services.
Innovation Solution
A method involving mining alarm data using frequent itemsets and machine learning to extract feature parameters, self-defining classification types, and determining classification weight values to construct an alarm root-cause relational tree model, which leverages traditional expert knowledge bases to adapt to new network topologies and services.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional expert knowledge bases are used to determine alarm root-cause relations, then the system has mature knowledge for traditional networks, but it cannot adapt to new 5G network topologies and services
Solution Approach 1:
The patent transforms the static expert knowledge base into a dynamic system that automatically learns and updates alarm root-cause relations. The system continuously mines alarm data from new 5G networks, extracts frequent itemsets, and updates the knowledge base without manual intervention, enabling adaptation to evolving network topologies while maintaining manageable complexity
Solution Approach 2:
The system performs self-updating by automatically mining alarm data, extracting frequent itemsets, and refining the knowledge base without requiring expert manual input. This self-service mechanism enables the system to adapt to new 5G networks autonomously, resolving the contradiction between adaptability and updating complexity
2Productivity
If alarm data is mined in traditional networks with mature expert knowledge bases, then data mining can be performed, but the loss outweighs the gain and there is not enough incentive
Solution Approach 1:
The patent performs preliminary action by building a comprehensive alarm root-cause knowledge base before it becomes obsolete. By proactively mining data from traditional networks and preparing the knowledge base for 5G transitions, the system ensures continuous relevance and value, making data mining worthwhile rather than a loss
Solution Approach 2:
The system implements feedback loops where mined alarm data from traditional networks is used to update and refine the knowledge base. This continuous feedback ensures that the mining process generates tangible improvements to the system's diagnostic capabilities, creating a positive cycle where the gain from mining consistently outweighs the cost
3Measurement precision
If complex factors such as topological relation, service relation, alarm severity and expert experience are used to determine root-cause relations, then the accuracy is improved, but the system complexity increases
Solution Approach 1:
The patent segments the complex root-cause determination process into distinct analytical components: topological relation analysis, service relation analysis, alarm severity assessment, and frequent itemset mining. Each component processes specific aspects independently, then integrates results to achieve high accuracy without overwhelming system complexity
Solution Approach 2:
The patent replaces manual expert judgment with automated data mining algorithms and machine learning models. The system objectively analyzes alarm data patterns, topological relationships, and service dependencies using computational methods, eliminating subjective variability while maintaining high accuracy and reducing operational complexity
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
A method for constructing a cloud network alarm root cause relational tree model, a device, and a storage medium. The method comprises: mining, according to a frequent itemset, first alarm data of a traditional network to obtain a first alarm tree model, and performing, by using a traditional expert knowledge database as a label, feature extraction on the first alarm tree model to obtain a first feature parameter; customizing, according to a pre-determined classification rule, classification types, and determining, by means of machine learning, a classification weight value corresponding to each of the classification types in the first alarm tree model; and mining, according to the frequent itemset, second alarm data of a network awaiting model-building, and performing, by using machine learning and the classification weight value, model-building on the second alarm data to obtain a second alarm tree model, the second alarm data being root-cause relation data.