一种基于多维数据的异常行为识别方法、介质及设备
By dividing and dynamically aggregating time periods based on the timeliness of data dimensions, and combining mutation index and sparsity for feature extraction, the problem of untimely identification of abnormal behavior in existing technologies is solved, achieving higher identification accuracy and early warning efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HANGZHOU YUNSHEN TECH CO LTD
- Filing Date
- 2026-04-16
- Publication Date
- 2026-07-17
AI Technical Summary
Existing methods for identifying abnormal behavior ignore the timeliness differences across different data dimensions. This leads to the inability to capture short-term mutation signals in a timely manner when device behavior changes drastically or when abnormal behavior is in a long incubation period. These signals are either diluted by historical data, resulting in untimely warnings or an increased false alarm rate.
Based on the degree of correlation between each data dimension and historical abnormal behavior, the system is divided into high, medium, and low timeliness types, dynamically determines the aggregation time period, and extracts features through mutation index, abnormal precursors, and sparsity to generate a high-quality feature dataset that adapts to the dynamic changes in device behavior.
It improves the accuracy of abnormal behavior identification, reduces the false negative rate, ensures the timeliness and accuracy of early warning, and balances data integrity and timeliness.
Smart Images
Figure CN122020499B_ABST