Correlation Function Selection for Abnormality Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing invariant relation analysis methods generate correlation functions with low abnormality detection ability, leading to undetectable abnormalities in metrics due to small prediction error thresholds, especially when abnormalities in non-objective or objective metrics have minimal impact on prediction values.
Innovation Solution
A system analysis device and method that extracts a correlation function from candidates based on detection sensitivity, which indicates the likelihood of correlation destruction during metric abnormalities, to enhance abnormality detection capability in invariant relation analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a correlation function is determined based on prediction accuracy (fitness), then the prediction accuracy is maximized, but the abnormality detection ability becomes low when the effect of abnormality on prediction value is small
Solution Approach 1:
The patent changes the selection criterion for correlation functions from prediction accuracy (fitness) to detection sensitivity. This parameter change allows the system to select correlation functions that are more responsive to abnormalities, even if they have slightly lower prediction accuracy. The detection sensitivity is calculated based on the magnitude of prediction errors when abnormalities occur, ensuring that the selected correlation function will generate detectable correlation destruction when metric abnormalities occur.
2Stability of the object's composition
If the effect of abnormality on prediction value is small, then the prediction error remains below threshold, but the abnormality becomes undetectable
Solution Approach 1:
The patent performs preliminary calculation of detection sensitivity for each candidate correlation function before selecting the final correlation function. By calculating the expected prediction error magnitude under abnormal conditions in advance, the system can identify and select correlation functions that will produce detectable errors when abnormalities occur. This preliminary action ensures that the selected correlation function has high abnormality detectability while maintaining acceptable prediction accuracy under normal conditions.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
In invariant relation analysis, a correlation model having high abnormality detection ability is generated. A system analysis device (100) includes a correlation function storage unit (1022), and a correlation function extraction unit (1023). The correlation function storage unit (1022) stores a plurality of candidates for a correlation function expressing a correlation for each pair of metrics in a system. The correlation function extraction unit (1023) extracts a correlation function from a plurality of candidates for a correlation function as a correlation function for each pair of metrics, on the basis of a detection sensitivity indicating a likelihood of correlation destruction caused at the time of abnormality of a metric associated with a correlation function.