Runtime Preemptive Alerts Using Adaptive Factor Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems struggle to provide preemptive alerts for potential computer system and application problems during runtime, especially in distributed and multi-cloud environments where data volumes are large, varied, and overwhelming, often masking underlying issues.
Innovation Solution
A method involving the selection of multiple sets of candidate factors, training classification models, validating them with datasets, adjusting factor weights, and iteratively improving models by replacing low-accuracy factors with high-accuracy ones using transformers and classification models to predict future system problems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple classification models are trained with different candidate factors to improve prediction accuracy, then the accuracy of preemptive alerts is improved, but the device complexity and computational resources required increase
Solution Approach 1:
The patent segments the candidate factors into multiple distinct sets, where each set is used to train a separate classification model. This segmentation allows the system to explore different factor combinations systematically, improving prediction accuracy by selecting the best-performing model while managing complexity through structured organization of training data and models.
Solution Approach 2:
The patent changes parameters by adjusting the composition of candidate factor sets and using accuracy scores to dynamically select and update models. The system modifies the input parameters (candidate factors) and model configurations based on performance metrics, enabling adaptive improvement of prediction accuracy while controlling complexity through parameter optimization.
2Reliability
If factor weights are adjusted based on accuracy scores to improve model performance, then the reliability of predictions is improved, but the time and computational resources for model validation increase
Solution Approach 1:
The patent applies preliminary action by pre-adjusting factor weights based on accuracy scores obtained during the training and validation process. Instead of performing extensive validation for each weight adjustment, the system uses accuracy scores from initial model training to guide weight adjustments, thereby improving reliability while reducing the time required for iterative validation.
Solution Approach 2:
The system implements feedback mechanisms where accuracy scores from model validation are used to adjust factor weights and improve subsequent predictions. This feedback loop allows the system to learn from validation results and automatically optimize model parameters, improving reliability while minimizing the need for repeated full validations.
3Productivity
If low accuracy classification models are updated by replacing candidate factors with higher performing ones, then the productivity of the alert system is improved, but the measurement precision may temporarily decrease during the update process
Solution Approach 1:
The patent applies dynamics by making the candidate factor sets and model configurations adaptable and updateable. The system dynamically replaces underperforming candidate factors with higher-performing ones based on accuracy comparisons, allowing the alert system to improve productivity over time while managing precision through controlled updates and validation processes.
Data Source
AI summary
For a set in sets of candidate factors, a classification model is trained to predict a computer problem possibility, an accuracy score of the classification model is determined based on model validation, and factor weights of the candidate factors in the set are adjusted based on the accuracy score. This processing is done with respect to all sets of candidate factors. A low accuracy classification model having an accuracy score lower than a threshold criterion is selected. A higher accuracy classification model having an accuracy score that is higher than the accuracy score of the low accuracy classification model is selected. The set of candidate factors used to train the low accuracy classification model is updated using one or more of the candidate factors used to train the higher accuracy classification model. The low accuracy classification model is updated based on the updated set of candidate factors.