Anomaly Detection in Multi-Dimensional Time Series
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Analyzing multi-dimensional time series data for anomalies is inefficient due to the large number of dimensions and their possible values, requiring extensive analysis and different detection techniques for each time series, which is computationally difficult and time-consuming.
Innovation Solution
A system that selects and applies an anomaly detection technique from a plurality of techniques to time-series data sets for different combinations of dimension values, reducing the number of time series that need to be analyzed by iteratively adding dimensions and eliminating non-anomalous dimension values, thereby focusing on the source of anomalies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If anomaly detection is performed on all time-series data sets for all combinations of dimension values, then anomaly detection completeness is improved, but processing time and computational resources increase significantly
Solution Approach 1:
The patent segments the anomaly detection process into multiple stages by dividing dimensions into two sets: a first set of dimensions for initial analysis and a second set of dimensions for further analysis. This segmentation allows the system to process data in manageable chunks, analyzing anomalies in the first dimension set before expanding to the second dimension set, thereby reducing overall processing time while maintaining detection completeness.
Solution Approach 2:
The patent introduces a dimensional approach where anomalies are first detected in a reduced dimensionality space (first set of dimensions) and then investigated in higher dimensionality (second set of dimensions). This dimensional progression allows the system to identify potential anomalies efficiently in lower dimensions before performing more computationally intensive analysis in higher dimensions, resolving the contradiction between completeness and processing time.
2Measurement precision
If multiple anomaly detection techniques are applied to each time series, then detection accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent applies different anomaly detection techniques selectively based on the specific dimension being analyzed. Instead of uniformly applying all detection techniques to all time series, the system chooses appropriate techniques for specific dimensions, optimizing detection accuracy for each dimension while reducing overall computational complexity by avoiding unnecessary technique applications.
Solution Approach 2:
The patent employs partial action by applying anomaly detection techniques only to the necessary extent. The system analyzes the first set of dimensions with initial detection techniques and only applies additional techniques to the second set of dimensions when anomalies are detected in the first set, avoiding excessive computational actions on data that does not require deep analysis.
3Adaptability or versatility
If the number of dimensions and dimension values is increased, then anomaly detection coverage is improved, but the number of time-series data sets to analyze increases exponentially
Solution Approach 1:
The patent segments the dimensions into two distinct sets, analyzing the first set independently before incorporating the second set. This segmentation prevents the system from needing to process all possible combinations of dimensions simultaneously, thereby controlling the quantity of time-series data sets that must be analyzed while maintaining comprehensive anomaly detection coverage across all dimensions.
Solution Approach 2:
The patent manages the exponential growth of data sets by transitioning through dimensional stages. First, the system analyzes time-series data for the first set of dimensions, creating a manageable initial quantity of data sets. Then, upon detecting anomalies, it extends analysis to the second set of dimensions, effectively managing the quantity of data sets at each dimensional stage rather than confronting the full exponential quantity at once.
Data Source
AI summary
Methods, systems, apparatuses, and computer program products are described herein that enable detecting anomalies in time series. An anomaly detection technique is selected from a plurality of detection techniques, and is applied to a first time-series data set (having a first set of dimensions). In response to detecting an anomaly in the time-series data set, the anomaly detection technique is applied to a second time-series data set that is a subset of the first time-series data set. The first time-series data set includes the first set of dimensions plus one or more additional dimensions.