IT Event Embedding Vector via Graph Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional IT systems struggle to analyze and group heterogeneous and asynchronous IT operations event data effectively, leading to inefficiencies in identifying and resolving incidents, which can be costly for organizations.
Innovation Solution
A method and system utilizing a graph database and embedding model, including a clustering algorithm and transformer with graph attention network encoder, to determine a signature embedding vector for IT events, enabling efficient analysis and correlation of asynchronous event data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional data organization techniques are applied to IT operations event data, then the system structure remains simple, but the ability to analyze and group heterogeneous and asynchronous event data deteriorates
Solution Approach 1:
The patent introduces an embedding model as an intermediary layer between raw IT event data and analysis systems. This embedding model transforms heterogeneous and asynchronous event data into standardized embedding vectors, enabling effective analysis without directly complicating the overall system architecture. The embedding vectors serve as a mediator that captures complex patterns while maintaining system modularity.
Solution Approach 2:
The patent transforms event data from its original heterogeneous format into embedding vectors through parameter transformation. By changing the representation parameters of event data (from raw logs to embedding vectors), the system gains improved analysis capability while the transformation process is handled by the specialized embedding model, isolating complexity.
2Measurement precision
If more detailed event data is collected for better incident analysis, then the precision of incident identification improves, but the time and resources required for processing increase
Solution Approach 1:
The patent applies preliminary action by pre-processing event data through the embedding model to create embedding vectors in advance. This preliminary transformation organizes and structures the data before actual incident analysis occurs, so that when incidents need to be identified, the processing is faster and more efficient, reducing the time loss during critical analysis phases.
Solution Approach 2:
The embedding model creates a compressed representation (copy) of the detailed event data. Instead of processing all raw event data during incident analysis, the system works with the embedding vector copies that capture the essential patterns, significantly reducing processing time while maintaining identification precision.
3Adaptability or versatility
If heterogeneous event data from multiple sources is integrated, then the comprehensiveness of incident analysis improves, but the complexity of data grouping and correlation worsens
Solution Approach 1:
The embedding model serves as a universal processor that handles multiple types of heterogeneous event data from different sources through a single unified transformation process. This multi-functional approach allows the system to integrate diverse data sources without creating separate processing paths, reducing the overall complexity of data grouping and correlation while improving integration capability.
Data Source
AI summary
A computer-implemented method for determining a signature embedding vector for information technology events. The method including: receiving a current data object representing an occurrence of an information technology event associated with a configurable item; obtaining a graph database of logical associations of the configurable item; applying a clustering algorithm to the graph database to determine a subset of graphs; extracting associated data of the subset of graphs; applying an embedding model to the extracted associated data to determine an embedding vector for each subset of graphs; aggregating the embedding vector for each subset of graphs to determine a signature embedding vector; and utilizing the signature embedding vector to perform further analysis of the current data object.