Contextual Anomaly Detection Across IoT Assets
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing anomaly detection techniques in IoT systems rely on supervised machine learning and fail to capture the temporal structure of time-series data, leading to inefficiencies in detecting and remediating issues across assets.
Innovation Solution
The method employs time-series segmentation, graphical Gaussian modeling, unsupervised graph clustering, and Sparse Mixture of Sparse Gaussians to cluster assets based on temporal context, identifying anomalous assets and cohorts through context-window distribution learning and isolation metrics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If supervised machine learning techniques are used for anomaly detection, then detection accuracy may be improved, but the ability to capture temporal structure of time-series data deteriorates
Solution Approach 1:
The patent segments time-series data into contextual windows, dividing the continuous temporal data into manageable segments that can be analyzed while preserving their temporal relationships. This segmentation allows the system to maintain temporal structure information while performing anomaly detection on segmented data frames.
Solution Approach 2:
The patent transforms temporal data into a different dimensional representation by creating contextual windows and cohorts, adding a cohort dimension to the temporal dimension. This dimensional transformation enables the system to capture temporal structures in a new space where anomaly detection can be performed effectively.
2Productivity
If unsupervised anomaly detection is implemented across multiple assets, then detection speed is improved, but the complexity of analyzing temporal contexts deteriorates
Solution Approach 1:
The patent merges multiple assets into cohorts based on their temporal patterns and characteristics. By grouping similar assets together, the system reduces the complexity of analyzing each asset individually while maintaining the ability to detect anomalies across the entire fleet. This merging approach enables faster unsupervised anomaly detection.
Solution Approach 2:
The patent creates a universal anomaly detection framework that works across multiple asset types and contexts. The cohort-based approach provides a multi-functional system that can detect anomalies at both the individual asset level and the cohort level, reducing overall system complexity through a unified detection mechanism.
3Measurement precision
If temporal context windows are applied to each asset, then anomaly detection precision is improved, but computational complexity deteriorates
Solution Approach 1:
The patent segments the computational task by applying temporal context windows to divide time-series data into manageable segments. This segmentation reduces the computational burden by processing smaller windows of data rather than analyzing entire time-series, while still maintaining detection precision through the preservation of temporal relationships within each window.
Solution Approach 2:
The patent applies partial action by focusing computational resources on analyzing temporal contexts only for assets or cohorts that exhibit anomalous behavior. Rather than performing exhaustive temporal analysis on all assets equally, the system applies temporal context windows selectively to reduce overall computational complexity while maintaining high detection precision for anomalous cases.
Data Source
AI summary
Methods, systems, and computer program products for contextual anomaly detection across assets are provided herein. A method includes obtaining time-series data frames corresponding to assets; clustering the assets into one or more cohorts based on the time-series data frames, each cohort comprising assets having statistically similar time-series data frames; for each given asset within each cohort: applying a time-context window to the portion of the time-series data frames corresponding to the given asset to generate at least one transformed data frame, and determining an asset distribution for the given asset based on the at least one transformed data frame; determining one or more of that at least one of the assets within at least one of the cohorts is anomalous and that at least one of the cohorts is anomalous; and causing at least one remediation action to be performed based on the determining.