ML Root Cause Analysis Using Historical Incident Action Data

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current root cause analysis systems for computing incidents are inefficient, often leading to delayed corrective actions, increased incident severity, and require excessive manual effort, resulting in inconsistent and time-consuming analyses.

Innovation Solution

A machine learning-based approach that trains a model using historical incident data, root cause data, and action data to quickly identify and automate root cause analysis, reducing manual interactions and improving accuracy and speed of incident response.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual root cause analysis methods are used, then human expertise can be applied to complex incidents, but the time and effort required for analysis increases significantly

Engineering Contradiction:

Improveaccuracy of root cause identificationVSAvoidtime required for root cause analysis

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent introduces a machine learning model as an intermediary between incident data and root cause identification. The model is trained on historical incident data, version history, and root cause information to automatically analyze new incidents and identify potential root causes, thereby reducing manual analysis time while maintaining accuracy through learned patterns from historical data

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system performs preliminary actions by pre-training the machine learning model on extensive historical incident data, version history, and root cause information before actual incident analysis. This preliminary training enables the model to quickly identify root causes during actual incidents without requiring manual analysis of the entire historical dataset each time

Inventive Principle:

Principle #10Preliminary action

2Measurement precision

If manual analysis methods are used, then detailed human inspection can be performed, but the number of devices and computational resources required increases

Engineering Contradiction:

Improveaccuracy of incident analysisVSAvoidnumber of devices needed for analysis

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent creates a virtual copy of the root cause analysis capability through a machine learning model that replicates human expert knowledge. Instead of requiring multiple human analysts or complex manual inspection systems, the model captures and reproduces analytical expertise in software, reducing the need for additional devices and computational infrastructure

Inventive Principle:

Principle #26Copying

3Productivity

If rapid incident analysis is implemented, then downtime is reduced, but the accuracy of root cause identification may be compromised

Engineering Contradiction:

Improvespeed of incident resolutionVSAvoidaccuracy of root cause determination

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The system performs preliminary training of the machine learning model on comprehensive historical incident data, version history, and root cause information before actual incident analysis. This advance preparation enables the model to quickly and accurately identify root causes during actual incidents without compromising accuracy for speed

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system incorporates feedback mechanisms where the machine learning model's predictions are validated against actual root causes from historical data, and the model is continuously retrained and improved. This feedback loop ensures that rapid analysis maintains high accuracy by learning from past performance and correcting errors

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS12632328B2Computer-based systems and/or computing devices configured for root cause analysis of computing incidents using machine learning based on user generated root cause analysis data of historical computing incidents

Publication Date: 2026.05.19 CAPITAL ONE SERVICES LLC

AI summary

An example method includes receiving incident data for historical incidents of downtime or interrupted service. The incident data includes identification information about one or more first computing applications, devices, or services affected by the downtime or the interrupted service. The incident data further includes timing information relating to the historical incidents and version history information of the one or more first computing applications, devices, or services. The method further includes receiving root cause data indicating a cause of the historical incidents and receiving action data indicating a corrective or preventative action taken or to be taken in response to each of the historical incidents. The method further includes training a machine learning algorithm using the incident, root cause, and action data to create a trained model configured to determine a root cause and a new corrective or preventative action for a new incident.