Bi-Directional Anomaly Detection for Sparse Time Series
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing anomaly detection techniques in computing systems are prone to high false positives and false negatives, failing to accurately identify anomalies in time series data, especially in sparse datasets and those accumulated over long periods.
Innovation Solution
A signed bi-directional anomaly detection algorithm that compares a given data item's value to its temporal neighborhoods from both earlier and later times, using quartile values to determine anomalies, allowing for more precise detection by considering local fluctuations and trends.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing anomaly detection techniques are used, then detection coverage is achieved, but false positives and false negatives increase
Solution Approach 1:
The patent segments the time series data into local neighborhoods (first neighborhood from earlier data items, second neighborhood from later data items) and performs separate anomaly detection for each neighborhood. This segmentation allows the algorithm to capture local fluctuations and trends independently, improving detection precision while reducing false positives and negatives by comparing each data item against its specific local context rather than using global thresholds.
2Measurement precision
If traditional anomaly detection methods are applied, then processing speed is maintained, but detection precision decreases
Solution Approach 1:
The patent performs preliminary actions by pre-calculating and storing quartile values (first quartile and third quartile) for each local neighborhood before anomaly detection. This preliminary computation of statistical parameters enables the actual anomaly detection to proceed efficiently by simply comparing data items against pre-computed thresholds, thereby improving detection precision without significantly compromising processing speed.
3Device complexity
If simple anomaly detection algorithms are used, then computational complexity is reduced, but detection capability in sparse and long-term data deteriorates
Solution Approach 1:
The patent applies local quality by making the anomaly detection algorithm adaptive to local characteristics of different data regions. Instead of using a uniform detection threshold across the entire time series, the algorithm computes separate first and third quartile values for each local neighborhood, allowing the detection criteria to vary according to local data density and patterns. This enables effective anomaly detection in both sparse and long-term data regions while maintaining relatively simple computational complexity.
Data Source
AI summary
Aspects of the present disclosure relate to systems and protocols for detecting anomalies in time series data. Embodiments include comparing the value for a data item to values for a neighborhood of data items from before the data item in the time series and values for a neighborhood of data items from after the data item in the time series. Based on the comparisons, an anomaly may be detected using an anomaly detection algorithm. Furthermore, the anomaly detection protocol may be integrated into computing systems to detect anomalous activity and performance metrics. Based on the detected anomalies, remedial actions may be performed with respect to the computing systems.