ML Model for Software Defect Escalation Prediction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Software developers face challenges in prioritizing and resolving software defects before release, as it is not always possible to address all defects due to limited time and resources, and manually predicting the likelihood of defect escalation is difficult, especially in large codebases.

Innovation Solution

A machine learning model is trained using past software defect escalations from prior versions to predict the likelihood of escalation for current software defects, considering attributes such as code changes, developer involvement, and defect descriptions, to prioritize defect resolution.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If all software defects are resolved prior to release, then product quality and reliability are improved, but development time and resource requirements increase significantly

Engineering Contradiction:

Improvesoftware defect resolutionVSAvoiddevelopment time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The patent applies preliminary action by training a machine learning model on historical defect data before the release decision needs to be made. The model learns patterns from past defect escalations and is ready to predict likelihood of escalation for current defects, enabling prioritization without manual analysis of each defect's potential impact.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent replaces manual heuristic prediction of defect escalation with an automated machine learning system. Instead of developers manually assessing each defect's likelihood to escalate, the ML model automatically analyzes defect attributes and historical patterns to generate escalation likelihood predictions, substituting human judgment with computational analysis.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Measurement precision

If manual prediction of defect escalation likelihood is performed, then developer judgment is utilized, but accuracy is insufficient especially in large codebases

Engineering Contradiction:

Improvedefect escalation prediction accuracyVSAvoidcodebase size

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent substitutes manual developer judgment with a machine learning model that systematically analyzes defect attributes and historical escalation patterns. The ML model processes multiple defect characteristics (code changes, developer history, defect type) and compares them against learned patterns from historical data, providing more consistent and accurate predictions than manual assessment, especially in large codebases where human cognitive limits apply.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent incorporates feedback by training the machine learning model on historical defect escalation data. The model learns from past outcomes which defect attributes and patterns correlated with actual escalations, continuously improving its prediction accuracy. This feedback loop allows the system to adapt to the specific codebase and organizational patterns, enhancing measurement precision over time.

Inventive Principle:

Principle #23Feedback

3Productivity

If prioritization of defect resolution is implemented, then resource efficiency is improved, but determining priority requires accurate prediction capability

Engineering Contradiction:

Improvedefect resolution efficiencyVSAvoiddefect escalation likelihood

Core Design Contradiction:

ProductivityVSDifficulty of detecting and measuring

Solution Approach 1:

The patent replaces the difficult manual task of predicting defect escalation likelihood with an automated machine learning system. The ML model analyzes multiple defect attributes (code changes, developer involvement, defect type, product area) and historical patterns to generate quantitative escalation likelihood predictions, making the invisible concept of 'escalation risk' detectable and measurable through computational analysis rather than human intuition.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS12141048B2Machine learning model for determining software defect criticality

Publication Date: 2024.11.12 SERVICENOW INC

AI summary

A system includes a processor configured to perform operations, including determining, for each respective software defect of software defects identified in a software product, corresponding attribute values that represent a software development history of the respective software defect, and determining, for each respective defect, using a machine learning model, and based on the corresponding attribute values, a corresponding escalation value representing a likelihood of the respective defect being escalated for resolution after release of the software product. The machine learning model may have been trained using corresponding software development histories of training defects that were escalated for resolution after release of a prior version of the software product. The operations also include, based on the corresponding escalation value of each respective defect, selecting, for resolution prior to the release of the software product, a defect subset of the software defects, and storing a representation of the defect subset.