Vector-Based Event Grouping for Computing Fault Resolution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Computing system fault events are difficult to properly group and identify due to the wide array of parameters associated with each issue, leading to inefficient automation of issue resolution and inadequate notification to relevant parties, as existing systems struggle to classify and correlate similar problems effectively.
Innovation Solution
A computer-implemented method that generates vectors from historical system fault data using word embedding models, clusters similar events, removes duplicates based on similarity metrics, and creates cluster and group centroids to automate the grouping and solution generation for system fault events, thereby improving detection and classification without requiring additional computing resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional event classification methods are used, then the system can process events with basic parameters, but it fails to properly group events due to the wide array of parameters associated with each issue
Solution Approach 1:
The patent transforms event parameters from their original diverse formats into unified vector representations using word embedding models. This parameter transformation allows events with widely varying parameters to be compared and grouped based on their semantic meaning rather than exact parameter matches, resolving the classification accuracy issue while handling parameter complexity.
Solution Approach 2:
The patent introduces vector embeddings as an intermediary representation layer between the raw event parameters and the classification system. This intermediary transforms complex, diverse parameters into a standardized vector space where similarity can be measured, enabling accurate event grouping without being overwhelmed by parameter complexity.
2Productivity
If manual incident grouping is performed, then solutions can be found for specific incidents, but the process is inefficient and does not scale to automate issue resolution
Solution Approach 1:
The patent implements self-service automation where the system automatically groups new events with historical events and generates solution recommendations without human intervention. The vector-based similarity search and automated grouping eliminate manual classification efforts, dramatically improving productivity while reducing the time loss associated with manual incident processing.
Solution Approach 2:
The patent performs preliminary actions by pre-processing historical event data into vector representations and organizing them into grouped clusters before new events arrive. When new events occur, the system can quickly match them against pre-organized groups rather than manually classifying each event, significantly reducing classification time and improving resolution efficiency.
3Measurement precision
If events are grouped based on multiple parameters, then more accurate classification is achieved, but the system generates noise and misclassification due to the complexity of correlation
Solution Approach 1:
The patent merges multiple event parameters into unified vector representations that capture the essential semantic meaning of each event. By combining parameters into holistic vector embeddings rather than analyzing each parameter separately, the system achieves accurate grouping while avoiding the noise and misclassification that arise from complex multi-parameter correlation.
Solution Approach 2:
The patent creates vector copies of event parameters that preserve the semantic relationships between parameters while eliminating the complexity of the original parameter structures. These vector copies enable accurate similarity measurement without the harmful effects of parameter complexity, reducing classification noise while maintaining grouping accuracy.
Data Source
AI summary
A method, computer system, and a computer program for grouping a plurality of computing system fault events is provided. The present invention may include extracting a summary of computing system fault events based on at least one similarity detected. The present invention may then include generating a plurality of vectors in which each vector corresponds to a summary, clustering the plurality of vectors into a plurality of clusters based on the at least one similarity, and compressing each cluster of the plurality of clusters into at least one cluster centroid. The present invention may further include generating a group centroid for a group including the plurality of clusters based on the at least one cluster centroid. The present invention may also include presenting a correlation statement derived from a result associated with the group centroid and generating a system fault solution based on the correlation statement.