Sliding Time-Window Anomaly Alarms Across Period Boundaries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional data anomaly detection methods often miss alarms due to splitting statistics across time period boundaries, leading to untriggered alarms and resulting losses for enterprises.
Innovation Solution
A data anomaly statistics and alarm method that uses a sliding time window to count anomalous data and generates alarms based on preset thresholds, with the ability to move the window backward according to a stepping duration rule, ensuring accurate detection and alerting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If data anomaly detection uses fixed time period boundaries (natural days, weeks, months) for statistics, then the evaluation process is simple and clear, but alarms may be missed when anomalies occur near time period boundaries as statistics are split across periods
Solution Approach 1:
The patent transforms the static fixed time period boundary approach into a dynamic sliding time window approach. The time window can slide backward by a configurable stepping duration to re-evaluate data, allowing the detection period to adapt dynamically rather than being constrained by fixed natural boundaries. This resolves the contradiction by maintaining operational simplicity through configurable parameters while improving alarm detection accuracy through flexible re-evaluation.
Solution Approach 2:
The patent introduces configurable parameters including time window length and stepping duration that can be adjusted based on different data quality levels and alarm scenarios. By changing these temporal parameters, the system can optimize between simplicity and accuracy for different situations, resolving the contradiction through parameter adaptability rather than fixed structure.
2Reliability
If the time window length and stepping duration are increased to improve detection coverage, then more anomalies can be detected, but the detection response time and system resource consumption increase
Solution Approach 1:
The patent establishes a mapping between data quality levels and specific time window configurations. Different quality levels (first, second, third level) correspond to different time window lengths and stepping durations, allowing the system to automatically adjust detection parameters based on data importance. This resolves the contradiction by providing appropriate detection coverage for each quality level without uniformly increasing response time for all data.
Solution Approach 2:
The patent applies different time window configurations to different quality levels of data, treating critical data (first level) differently from less critical data (third level). This local differentiation allows optimized detection coverage for each data type without universally increasing system response time, resolving the contradiction through localized parameter optimization.
3Reliability
If multiple time window configurations are used for different quality levels of data, then detection accuracy for critical data is improved, but the device complexity and configuration management increase
Solution Approach 1:
The patent defines a structured quality level system with three levels, each mapped to specific time window and stepping duration configurations. This standardized parameter mapping simplifies configuration management compared to arbitrary complex rules, as the system automatically selects appropriate parameters based on data quality level classification.
Solution Approach 2:
The patent segments data into three quality levels with distinct detection configurations, allowing differentiated handling of critical versus non-critical data. This segmentation enables focused optimization for critical data while maintaining manageable complexity through clear categorization and standardized configuration sets for each segment.
Data Source
AI summary
The present disclosure relates to the technical field of data anomaly detection and alarm, and more particularly, to a data anomaly statistics and alarm method and apparatus, and an electronic device, which effectively relieve the problem of alarm missing in the related technologies, thereby effectively avoiding the loss of enterprises due to alarm missing, the method includes: acquiring a detection time and a detection result of each of a plurality of data respectively through detecting; counting the number of target data in a current time window, and generating an alarm signal to prompt in response to the number of target data obtained by the counting being greater than a preset number threshold corresponding to a quality level of the data; moving the current time window backward according to a stepping duration included in an obtained stepping duration setting rule corresponding to the quality level of the data, and the current time window moved backward is used as a new current time window.