Anomaly Detection Model Using Grouped Time Series Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional anomaly detection methods in IT systems face challenges such as algorithm selection difficulty and low accuracy due to changing fluctuation modes in data sequences, limiting their effectiveness across different systems and indicators.
Innovation Solution
A system monitoring method that extracts multiple features from time series data, groups them into feature combinations, and uses these combinations as input for anomaly detection models, allowing for adaptive detection across various systems and indicators without relying on a single algorithm.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a single anomaly detection algorithm is used, then the detection process is simple, but the detection accuracy decreases when fluctuation modes change
Solution Approach 1:
The patent segments the anomaly detection process by dividing fluctuation modes into multiple categories (seasonal, trend, periodic, random) and selecting specific detection algorithms for each mode. This segmentation allows the system to achieve high accuracy for each specific mode while maintaining overall system simplicity through automated mode identification and algorithm selection.
Solution Approach 2:
The patent changes the parameter of algorithm selection based on the detected fluctuation mode. By dynamically adjusting which algorithm is applied according to the characteristics of the data (identified through statistical analysis and pattern recognition), the system maintains simplicity of operation while achieving high detection accuracy across varying conditions.
2Measurement precision
If multiple anomaly detection algorithms are used to handle different fluctuation modes, then the detection accuracy improves, but the algorithm selection difficulty increases
Solution Approach 1:
The patent implements self-service by enabling the system to automatically identify the fluctuation mode through statistical analysis and automatically select the appropriate detection algorithm without requiring user intervention. The system performs self-diagnosis of data characteristics and self-configures the detection approach, thereby improving accuracy while eliminating the complexity of manual algorithm selection.
Solution Approach 2:
The patent employs feedback mechanisms where the results of statistical analysis and preliminary pattern recognition feed into the algorithm selection process. The system continuously monitors data characteristics and adjusts the detection algorithm based on feedback about the actual fluctuation mode observed, thereby achieving high accuracy while keeping the selection process automated and manageable.
3Measurement precision
If anomaly detection algorithms are optimized for specific fluctuation modes, then the detection accuracy for that mode improves, but the adaptability to other modes decreases
Solution Approach 1:
The patent achieves universality by creating a multi-functional detection system that can handle multiple fluctuation modes (seasonal, trend, periodic, random) using a unified framework. The system incorporates multiple specialized algorithms within a single platform, automatically selecting the appropriate one based on the detected mode, thereby maintaining high accuracy for each specific mode while ensuring broad adaptability across different IT systems and indicators.
Data Source
AI summary
In a system monitoring method, a monitoring server extracts a plurality of features based on a plurality of data slices corresponding to a to-be-detected time period, where data of one of the data slices includes data in the to-be-detected time period. The monitoring server groups the plurality of features into a plurality of feature groups, and uses the plurality of feature groups as input data of an anomaly detection model. The monitoring server determines, based on a detection result of the anomaly detection model, whether the monitored indicator is abnormal.