Dynamic Graph Feature Extraction for Time Series Anomaly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Time series anomaly detection algorithms are unable to process graph data directly, making it difficult to detect anomalies in dynamic graphs, which are common in real-world applications such as social networks and pandemic spread analysis.
Innovation Solution
An incremental approach is developed to build and track connected components of dynamic graphs over time, converting the graph data into a time series dataset that can be processed by existing time series anomaly detection algorithms, allowing for efficient processing and anomaly detection in dynamic graph systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If time series anomaly detection algorithms are used, then anomaly detection capability is improved, but the ability to process graph data is lost
Solution Approach 1:
The patent introduces an intermediary transformation process that converts graph data into time series data format. This mediator enables compatibility between graph data structures and time series anomaly detection algorithms without requiring modification of the algorithms themselves. The transformation extracts temporal features from graph data and represents them in a format suitable for time series analysis.
Solution Approach 2:
The patent transforms the data representation parameters by converting graph-based parameters (nodes, edges, connections) into time series parameters (temporal sequences, numerical values). This parameter transformation allows the same anomaly detection algorithms to operate on both graph data and traditional time series data, resolving the format compatibility issue.
2Stability of the object's composition
If graph data is processed directly, then data structure integrity is improved, but compatibility with existing algorithms is worsened
Solution Approach 1:
The patent segments the graph data processing into distinct stages: graph analysis phase (maintaining structural integrity) and transformation phase (converting to time series format). This segmentation allows each stage to optimize for its specific requirement - structural preservation during analysis and format compatibility during transformation - without compromising either aspect.
3Measurement precision
If dynamic graphs are analyzed over time, then anomaly detection accuracy is improved, but computational complexity is worsened
Solution Approach 1:
The patent performs preliminary actions by pre-processing graph data to extract temporal patterns and transform them into time series format before applying anomaly detection algorithms. This preliminary transformation simplifies the subsequent detection process and reduces computational complexity during the actual anomaly detection phase, while maintaining high accuracy through preserved temporal information.
Data Source
AI summary
A plurality of graph snapshots for a plurality of consecutive periodic time samples maps between connected components in consecutive graph snapshots and describes at least one feature of each connected component. A recursively-built tree tracks an evolution of one of the connected components through the plurality of graph snapshots, the tree including a root node representing the connected component at a final one of the consecutive periodic time samples and a plurality of leaf nodes branching from the root node. A plurality of paths is extracted from the tree by traversing the tree from the root node to respective ones of the plurality of leaf nodes. Each path contains data describing an evolution of a respective one of the connected components through time as indicated by evolution of the at least one feature of the respective one of the connected components. Each of the plurality of paths is converted into a respective numerical vector of a plurality of numerical vectors that may be used as inputs to a time series anomaly detection algorithm.