Visual Anomaly Detection Model for APM Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing anomaly detection techniques in Application Performance Management (APM) rely on numerical methods that lack accuracy in capturing human judgment, making them inefficient in detecting anomalies, especially in multivariate cases, as they require complex and tedious processes to incorporate human judgment.
Innovation Solution
A method that visually generates and modifies an anomaly detection image using a coordinate system with distinguishable areas, allowing users to interactively define and modify these areas, which are associated with specific formulas, enabling the incorporation of human judgment into the anomaly detection process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If numerical methods are used for anomaly detection, then the detection process can be automated, but the accuracy in capturing human judgment deteriorates
Solution Approach 1:
The patent introduces an intermediary visual interface (coordinate system with distinguishable areas) that mediates between automated numerical methods and human judgment. Users interact with the visual representation to define anomaly regions, which are then converted into numerical formulas for automated detection, thus bridging the gap between automation and human expertise
Solution Approach 2:
The system performs preliminary visual setup where users define distinguishable areas and associated formulas before actual anomaly detection. This preliminary configuration captures human judgment in advance, allowing the subsequent automated detection process to leverage these pre-defined human-expertise-based parameters
2Measurement precision
If complex numerical methods are used to incorporate human judgment, then detection accuracy may improve, but the process complexity and tediousness increase
Solution Approach 1:
The patent replaces complex mechanical/n numerical processes with a visual system. Instead of requiring users to manually configure complex numerical parameters and formulas, the system allows interaction through visual manipulation of distinguishable areas on a coordinate system, which are then automatically converted into detection formulas
Solution Approach 2:
The system changes the parameter representation from complex numerical configurations to visual spatial parameters. Users define anomaly detection criteria through visual distinguishable areas in a coordinate system, and the system automatically transforms these visual parameters into the necessary numerical formulas for detection
3Stability of the object's composition
If traditional numerical methods are used, then the detection process can be standardized, but the ability to detect anomalies in multivariate cases deteriorates
Solution Approach 1:
The patent transitions from traditional one-dimensional or simple multi-dimensional numerical analysis to a visual two-dimensional coordinate system representation. This dimensional change allows better visualization and handling of multivariate data, where distinguishable areas can represent complex multivariate anomaly regions that are difficult to define with traditional numerical methods
Data Source
AI summary
A method includes formatting for display, on a visual screen, an image comprising: (1) a coordinate system, (2) a plurality of distinguishable areas within the coordinate system, each distinguishable area graphically representing a respective formula, and (3) a plurality of data points. The method also includes receiving user input comprising a modification to a particular distinguishable area. In response to receiving the user input, the method includes modifying one or more respective formulas based on the modification to the distinguishable area. For each data point, the method further includes associating the data point with one of the distinguishable areas by determining which of the modified formulas the data point falls within. The method further includes formatting for display a graphical representation of each modified formula. The method additionally includes storing the graphical representation of each modified formula for use as a modified image in future operations.