Temporal Incident Clustering for Faster Similar Event Resolution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing computing systems lack the ability to efficiently identify historically similar incidents, leading to prolonged incident resolution times and increased costs due to the inability to analyze and utilize previous incidents for quick resolution.
Innovation Solution
A computer-implemented method using temporal associations and machine learning models to cluster and associate similar incidents based on short descriptions, applying a frequency pattern growth algorithm to generate association rulesets for efficient incident analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If existing computing systems manually examine incidents without automated analysis, then system complexity remains low, but incident resolution time increases and productivity decreases
Solution Approach 1:
The system performs preliminary actions by pre-processing incident data to extract features, generate embeddings, and create association rulesets before actual incident resolution is needed. Historical incident data is continuously processed to build knowledge models that can be quickly queried when new incidents occur, eliminating the need for manual analysis during the resolution process.
Solution Approach 2:
The patent replaces manual mechanical analysis processes with automated machine learning models. Instead of human analysts manually examining incident descriptions and searching through historical data, the system uses neural networks to generate embeddings and association rules to automatically identify similar incidents, substituting human cognitive processes with automated computational mechanisms.
2Measurement precision
If the system stores and analyzes all historical incident data to find similar incidents, then incident resolution accuracy improves, but information processing time and resource consumption increase
Solution Approach 1:
The system extracts only the most relevant information from historical incident data by generating embeddings that capture essential features and characteristics. Instead of processing all raw historical data during incident resolution, the system extracts and stores condensed representations (embeddings) and pre-computed association rules, allowing quick retrieval of similar incidents without reprocessing the entire historical dataset.
Solution Approach 2:
The system performs preliminary processing of historical incident data to create pre-computed association rulesets and embedded representations. This preliminary action is performed once during data ingestion and continues to be updated, so that during actual incident resolution, the system can quickly query these pre-processed structures rather than analyzing raw historical data in real-time.
3Loss of time
If the system manually searches through historical incidents to find similar ones, then false positive rate decreases, but time consumption increases
Solution Approach 1:
The system incorporates feedback mechanisms where the quality and relevance of identified similar incidents are continuously evaluated and used to refine the association rulesets and embedding models. This feedback loop ensures that the automated identification becomes increasingly accurate over time, reducing false positives while maintaining fast processing speeds through continuous learning from actual incident resolution outcomes.
Data Source
AI summary
A computer-implemented method for determining related information technology event data by applying temporal associations includes: receiving a data object including a short description indicating an occurrence of a current incident associated with a configurable item; applying a first machine learning model to the short description to determine a first cluster associated with the data object; receiving a plurality of data object indicating occurrences of current incidents that occurred within a set period of time of the current incident, the plurality of data objects including a plurality of short descriptions; applying the first machine learning model to the short descriptions to determine associated clusters; determining, based on association rules and the associated clusters, a set of similar data objects from the plurality of data objects; assigning a set of associations between the data object and each of the set of similar data objects; and storing the set of associations.