Computing Environment Change Risk Scoring With Predictive Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining the risk of software and hardware updates in computing environments are complex and specific to incremental software updates, requiring extensive training and are unsuitable for hardware changes, leading to potential malfunctions and downtime.
Innovation Solution
A computer-implemented method using a trained predictive model that processes change features from a change record to calculate a risk score, flagging high-risk updates by analyzing historical data and learning relationships between change features and potential failures or malfunctions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing machine learning models are used to determine risk of software updates, then risk assessment capability is improved, but device complexity and training requirements increase significantly
Solution Approach 1:
The patent extracts only the essential risk-determining features from complex change requests and feeds them to a simplified machine learning model. By taking out only the necessary features (change type, timing, duration, impact scope) rather than processing entire change request documents, the system achieves effective risk assessment with reduced model complexity and training requirements.
Solution Approach 2:
The patent segments the risk assessment process into distinct components: feature extraction module, machine learning model, and risk scoring output. This segmentation allows each component to be optimized independently, with the ML model focusing only on predicting risk based on simplified features rather than processing complex code differences.
2Reliability
If existing machine learning models analyze code differences to determine risk, then software update risk assessment is improved, but adaptability to hardware changes is lost
Solution Approach 1:
The patent creates a universal risk assessment model that handles both software updates and hardware changes through a single unified approach. The machine learning model is trained on diverse change types and uses generalizable features (change type, timing, duration, impact) that apply across different domains, making the system versatile for various update scenarios without requiring domain-specific models.
3Measurement precision
If extensive training data is used to capture change risks, then risk prediction accuracy is improved, but processing time and computational resources increase
Solution Approach 1:
The patent extracts only the most relevant risk indicators from historical change data, such as change type, timing patterns, duration, and impact scope. By focusing training on these key features rather than processing entire historical change request documents, the system achieves accurate risk predictions with reduced training data volume and shorter training times.
Solution Approach 2:
The patent uses a representative sample of historical changes that captures the essential risk patterns without requiring complete analysis of all historical data. This partial action approach provides sufficient training accuracy while significantly reducing processing time and computational resources compared to exhaustive analysis.
4Measurement precision
If detailed analysis of change requests is performed, then risk identification accuracy is improved, but processing speed decreases
Solution Approach 1:
The patent extracts only the critical risk-determining features from change requests (change type, timing, duration, impact scope) and feeds these to the machine learning model. This extraction approach maintains high risk identification accuracy while dramatically improving processing speed compared to detailed analysis of entire change request documents.
Solution Approach 2:
The patent replaces manual or detailed mechanical analysis of change requests with an automated machine learning system that processes simplified features rapidly. The ML model automatically predicts risk scores based on extracted features, substituting slow detailed analysis with fast automated computation while maintaining or improving accuracy.
Data Source
Figure 1~2
Figure 3A~3B
Figure 4
AI summary
A computer-implemented method for flagging risk when handling updates to a computing environment is disclosed. The method includes receiving a change request comprising one or more change features associated with a change in hardware and/or software within the computing environment, and receiving a change record associated with the change request, the change record comprising one or more fields corresponding to the one or more change features. The method further includes calculating a risk score for the change by applying a trained predictive model to the one or more fields, the predictive model having been trained on a dataset comprising historical changes to the computing environment and their associated historical change records. If the calculated risk score is above a predetermined threshold, a flag is raised to the change request to notify a user