Nonlinear Sequence Analysis for Dynamic Anomaly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data mining techniques face challenges in effectively detecting anomalies in heterogeneous, multivariate data sets that vary as functions of one or more independent variables, as they often require complex computations and may not adequately capture non-linear relationships.
Innovation Solution
The method involves computing feature vectors at various values of independent variables using nonlinear sequence analysis, followed by detecting anomalies in how these output values change, employing techniques such as finite-time Lyapunov exponents, off-diagonal complexity, and temporal correlations, and optionally preprocessing the data to normalize or weight it.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If current data mining techniques are used to detect anomalies in heterogeneous, multivariate data sets, then anomaly detection capability is provided, but the computation becomes complex and non-linear relationships are not adequately captured
Solution Approach 1:
The patent transforms the anomaly detection problem by changing parameters from direct multivariate analysis to analyzing changes in output values as functions of independent variables. This involves computing feature vectors and transforming data through nonlinear sequence analysis, which simplifies the computational complexity while maintaining detection precision by focusing on temporal or sequential patterns in the transformed feature space
Solution Approach 2:
The patent replaces complex mechanical/computational data mining systems with a mathematical transformation approach using nonlinear sequence analysis. Instead of directly processing heterogeneous multivariate data through complex algorithms, the system substitutes this with a mathematical framework that computes feature vectors and analyzes their evolution, reducing computational complexity while capturing non-linear relationships
2Reliability
If complex computations are performed to detect anomalies in multivariate data sets, then detection thoroughness is improved, but processing time increases
Solution Approach 1:
The patent segments the complex multivariate anomaly detection problem into distinct computational stages: (1) computing feature vectors from input data, (2) transforming these vectors through nonlinear sequence analysis to obtain output values, and (3) detecting anomalies in the evolution of output values. This segmentation allows each stage to be optimized independently, improving detection thoroughness while managing processing time through modular computation
3Measurement precision
If traditional data mining methods are used, then existing anomaly patterns can be detected, but evolving clusters and uncharacteristic patterns are not effectively distinguished
Solution Approach 1:
The patent introduces dynamics by analyzing how output values change as functions of independent variables rather than examining static data points. The nonlinear sequence analysis computes how feature vectors evolve, allowing the system to detect anomalies in the temporal or sequential evolution of patterns. This dynamic approach enables effective distinction between evolving clusters and uncharacteristic patterns by capturing their behavioral differences over time or across variable values
Data Source
AI summary
Methods and systems for detecting anomalies in sets of data are disclosed, including: computing components of one or more types of feature vectors at a plurality of values of one or more independent variables, each type of the feature vectors characterizing a set of input data being dependent on the one or more independent variables; computing one or more types of output values corresponding to each type of feature vectors as a function of the one or more independent variables using a nonlinear sequence analysis method; and detecting anomalies in how the one or more types of output values change as functions of the one or more independent variables.