Time-Series Anomaly Detection via Data Window Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing anomaly detection systems cannot accurately identify periods in time-series data where data values are untrustworthy, leading to potential misinformed decision-making.
Innovation Solution
A method to detect an untrustworthy period in time-series data by identifying a start and end data window, using online detection and anomaly analysis to determine compensation status based on movement patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing anomaly detection systems are used to monitor time-series data, then abnormal events can be detected, but the system cannot accurately identify periods where data values are untrustworthy, leading to potential misinformed decision-making
Solution Approach 1:
The patent divides the time-series data into multiple data windows (e.g., 5-minute intervals) to segment the continuous data stream. This segmentation allows the system to identify specific time intervals where data becomes untrustworthy, rather than treating the entire time series as a single unit. By segmenting the data, the system can precisely locate and report untrustworthy periods, improving both measurement precision and data reliability assessment.
Solution Approach 2:
The patent performs preliminary anomaly detection to identify potential untrustworthy periods before final confirmation. The system continuously monitors data windows for anomalies and maintains a list of candidate untrustworthy periods. Only after verifying that the anomaly persists across multiple data windows does the system finalize the untrustworthy period identification. This preliminary action reduces false positives and improves the reliability of the identified periods.
2Productivity
If continuous monitoring of time-series data is performed, then real-time anomaly detection is achieved, but the complexity of determining compensation status and movement patterns increases
Solution Approach 1:
The patent segments the time-series data into fixed-time windows (e.g., 5-minute intervals), which simplifies the complexity of continuous monitoring. By working with discrete, fixed-size segments rather than continuous streams, the system can more easily compare movement patterns and determine compensation status. This segmentation maintains real-time processing capability while reducing computational complexity.
Solution Approach 2:
The patent transforms the monitoring approach by changing from analyzing individual data points to analyzing aggregate parameters within data windows. Instead of tracking every individual value change, the system monitors parameters such as the number of anomalies per window, the time span of untrustworthy periods, and the magnitude of data deviations. This parameter transformation simplifies the complexity of compensation status determination while maintaining real-time detection effectiveness.
Data Source
AI summary
The present disclosure proposes a method, apparatus and computer program product for detecting an untrustworthy period of a metric. Time-series data for a target metric may be obtained, the time-series data including a plurality of data windows. A start data window and an end data window of an untrustworthy period of the target metric may be identified from the time-series data, the untrustworthy period indicating a time interval in which data values of the target metric are untrustworthy. The untrustworthy period may be detected based on the start data window and the end data window.