Anomaly Detection in Multivariate Data Using Inferential Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing asset data platforms face challenges in detecting anomalies in multivariate observation vectors, particularly when these vectors contain variables with invalid values or interrelated variables, which can lead to unnecessary alerts and inaccurate predictions.
Innovation Solution
The use of inferential modeling in combination with component analysis allows the asset data platform to evaluate only a subset of variables in a multivariate observation vector, producing a predicted version with valid values for all variables, thereby improving anomaly detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional anomaly detection methods are used on multivariate data with invalid values or interrelated variables, then the system can process all variables uniformly, but the anomaly detection accuracy deteriorates due to false alerts and inaccurate predictions
Solution Approach 1:
The patent segments the multivariate observation vector into multiple groups based on variable relationships and validity status. Valid variables are separated from invalid variables, and interrelated variables are grouped together. This segmentation allows the system to process different variable groups using appropriate methods, improving anomaly detection accuracy while managing complexity through structured organization of the data.
2Reliability
If all variables in the observation vector are processed using the same method, then the processing is simple and uniform, but the detection reliability deteriorates when invalid values or interrelated variables are present
Solution Approach 1:
The patent applies local quality by using different processing methods for different variable groups. Valid variables undergo one type of processing while invalid variables undergo different processing. Similarly, interrelated variables are processed differently from independent variables. This localized approach to data processing improves detection reliability by tailoring the analysis method to the specific characteristics of each variable group, rather than applying a uniform method that would compromise reliability.
3Measurement precision
If the system processes multivariate data with invalid values using traditional methods, then the processing speed is maintained, but the prediction accuracy deteriorates due to unnecessary alerts
Solution Approach 1:
The patent performs preliminary actions by identifying and categorizing variables before the main anomaly detection process. Invalid variables are flagged and separated in advance, and interrelated variable groups are established beforehand. This preliminary organization allows the main processing to focus only on relevant variables, improving prediction accuracy by eliminating noise from invalid data while maintaining processing efficiency through pre-structured data organization that reduces computational overhead during the actual detection phase.
Data Source
AI summary
A data analytics platform may be configured to construct an inferential model for a multivariate observation vector using inferential modeling in combination with component analysis, which may enable the data analytics platform to evaluate only a subset of the variables in the observation vector and then output a predicted version of the multivariate observation vector that includes predicted values for the full set of variables that was originally included in the observation vector. In turn, the data analytics platform may use the predicted version of the multivariate observation vector output by the inferential model to determine whether an anomaly has occurred.