Service Hotspot Identification Using LLM Cluster Labeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
IT teams struggle to identify critical issues among a multitude of incoming requests and incidents in IT service management systems, leading to delays in addressing these issues, disruptions in business operations, and potential losses in productivity and revenue, due to the lack of a proactive framework for monitoring and analyzing IT infrastructure.
Innovation Solution
A method and system using a Large Language Model (LLM) to analyze raw data, generate embeddings, cluster them into sets of clusters and sub-clusters, and determine cluster labels to identify domain-specific service management hotspots, which are areas prone to recurrent issues or failure, by employing chunking, tokenizing, and similarity analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If IT teams manually analyze incoming requests and incidents, then they can identify critical issues, but the sheer volume and diversity of issues overwhelm the teams leading to delays
Solution Approach 1:
The patent segments the overwhelming volume of IT issues into manageable clusters using unsupervised machine learning algorithms. Similar incidents and requests are grouped together based on their characteristics, allowing the system to identify critical issue patterns efficiently without manually analyzing each individual item, thus resolving the contradiction between identification accuracy and resolution speed
Solution Approach 2:
The patent introduces an unsupervised machine learning model as an intermediary between raw IT service data and human analysts. This intermediary automatically processes and clusters the data, extracting meaningful patterns and presenting them in a digestible format, thereby enhancing both the speed and accuracy of critical issue identification without overburdening IT teams
2Reliability
If IT teams respond reactively to issues as they arise, then they can address immediate problems, but they miss opportunities to identify and mitigate hotspots proactively
Solution Approach 1:
The patent implements preliminary action by using unsupervised machine learning to continuously analyze and cluster IT service data in real-time, automatically identifying emerging patterns and potential hotspots before they manifest as full-blown incidents. This proactive clustering enables the system to alert teams to developing issues, allowing them to take preventive measures and mitigate problems before they disrupt service, thus improving reliability while reducing response time
3Difficulty of detecting and measuring
If no structured framework is used for analyzing IT infrastructure, then the system remains simple, but timely detection of emerging issues becomes challenging
Solution Approach 1:
The patent applies the self-service principle by implementing an unsupervised machine learning framework that automatically clusters and analyzes IT service data without requiring manual configuration or intervention. The system self-organizes the data based on inherent patterns, automatically identifying emerging issues and hotspots. This autonomous approach significantly enhances detection capability while keeping the operational complexity manageable, as the system adapts and learns from the data itself rather than requiring complex pre-defined rules
Data Source
AI summary
This disclosure relates to method and causal analysis server for identifying domain-specific service management hotspots. The method includes receiving raw data corresponding to plurality of domain-specific issues; generating plurality of chunks from the raw data using an optimal chunking and tokenizing algorithm selected from chunking and tokenizing algorithms to obtain chunked data; creating a plurality of embeddings from the chunked data through an embedding model; clustering the plurality of embeddings into a set of clusters and a set of sub-clusters associated with each of the set of clusters, based on a similarity analysis; determining a set of cluster labels corresponding to the set of clusters and a set of sub-cluster labels corresponding to the set of sub-clusters, via a Large Language Model (LLM); and identifying a plurality of hotspots corresponding to the plurality of domain-specific issues based on the set of cluster labels and the set of sub-cluster labels.