一种异常检测方法、装置及存储介质
By using Euclidean distance binary classification in an object ensemble classifier and bootstrap sampling in a support vector machine, the problem of anomaly detection in imbalanced datasets is solved, achieving data balance and information preservation, and improving the accuracy of anomaly detection for server monitoring metrics.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA TELECOM CLOUD TECH CO LTD
- Filing Date
- 2022-07-14
- Publication Date
- 2026-07-17
AI Technical Summary
When dealing with imbalanced datasets, existing technologies may use random undersampling techniques to remove samples with valuable information, resulting in poor training performance of subsequent algorithms. This is especially true in the detection of anomalies in server monitoring metrics, where normal samples far outnumber abnormal samples, leading to unsatisfactory data balancing.
A target ensemble classifier is adopted. The majority class samples are classified by Euclidean distance to determine the support vectors and perform bootstrap sampling. The base classifier is trained by combining the minority class sample subset to form a weighted ensemble classifier, which ensures the preservation of useful information and data balance.
It achieves good anomaly detection results on imbalanced datasets by removing a subset of majority class samples while retaining useful classification information, thus improving the accuracy and reliability of detection.
Smart Images
Figure CN115270939B_ABST