KPI Anomaly Detection via Historical Model Reuse
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Anomaly detection for key performance indicators (KPIs) is inefficient due to high time and resource consumption, especially as system complexity increases and data volumes grow, requiring manual analysis and repeated training of historical data for different KPIs.
Innovation Solution
A method and apparatus that determine a similar historical KPI to construct a model for a new KPI, using feature vectors and anomaly detection algorithms to identify optimal performance models, reducing the need for retraining and improving modeling efficiency by leveraging existing models for similar indicators.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If historical data is repeatedly trained for all different KPIs to obtain anomaly detection models, then anomaly detection accuracy is improved, but time consumption and resource consumption increase significantly
Solution Approach 1:
The patent pre-calculates and stores feature vectors for all historical KPI data in advance. When a new KPI needs anomaly detection, the system directly retrieves and compares feature vectors without repeating the entire training process, significantly reducing modeling time while maintaining detection accuracy
Solution Approach 2:
The patent copies the feature vector structure and anomaly detection logic from historical KPIs to new KPIs. By using the same feature extraction and comparison methodology across different KPIs, the system avoids retraining while preserving detection effectiveness
2Reliability
If historical data is repeatedly trained for all different KPIs to obtain anomaly detection models, then anomaly detection accuracy is improved, but resource consumption increases significantly
Solution Approach 1:
The system performs feature extraction and vector calculation in advance during data preprocessing, storing results for future use. This eliminates the need to recalculate features during each anomaly detection task, significantly reducing computational resource consumption
Solution Approach 2:
The patent reuses the same anomaly detection algorithm and feature comparison logic across different KPIs. By copying the proven detection mechanism rather than retraining models for each KPI, the system maintains accuracy while minimizing resource usage
3Reliability
If manual analysis is performed for anomaly detection, then detection accuracy can be maintained, but productivity decreases as data volume increases
Solution Approach 1:
The patent implements automated feature extraction and anomaly detection that operates independently without manual intervention. The system automatically compares new KPI data against historical patterns, maintaining professional-level detection accuracy while achieving high throughput and efficiency
Solution Approach 2:
The patent replaces manual analysis with automated computational methods. By substituting human experts with algorithmic feature comparison and anomaly detection, the system maintains detection quality while dramatically improving productivity and scalability
Data Source
AI summary
A modeling method and an apparatus are disclosed. The method includes: obtaining a first data set of a first indicator, and determining, based on the first data set, a second indicator similar to the first indicator; and determining a first model based on one or more second models associated with the second indicator. The first model is used to detect a status of the first indicator, and the status of the first indicator includes an abnormal state or a normal state. The second models are used to detect a status of the second indicator, and the status of the second indicator includes an abnormal state or a normal state.