Title:
DATA ABNORMALITY DETECTION METHOD AND APPARATUS
Document Type and Number:
WIPO Patent Application WO/2021/098384
Kind Code:
A1
Abstract:
A data abnormality detection method and apparatus, which are used to reduce resource overheads and meet real-time detection requirements. The method comprises: acquiring the current detection sample, a historical cell state and a historical hidden layer state of an object to be tested, wherein the historical cell state is an output cell state determined by means of inputting a historical detection sample of said object into an abnormality detection model of said object, and the historical hidden layer state is an output hidden layer state determined by means of inputting the historical detection sample of said object into the abnormality detection model of said object; determining the current feature value of said object according to the current detection sample; inputting the current feature value, the historical cell state and the historical hidden layer state of said object into the abnormality detection model to determine a risk index corresponding to said object; and comparing the risk index with a risk threshold value to determine an abnormality determination result for said object.
More Like This:
Inventors:
ZANG DAWEI (CN)
Application Number:
PCT/CN2020/118430
Publication Date:
May 27, 2021
Filing Date:
September 28, 2020
Export Citation:
Assignee:
CHINA UNIONPAY CO LTD (CN)
International Classes:
G06Q20/40; G06N3/04
Foreign References:
| CN110874744A | 2020-03-10 | |||
| CN110414666A | 2019-11-05 | |||
| CN109800858A | 2019-05-24 | |||
| CN107769972A | 2018-03-06 | |||
| CN110441065A | 2019-11-12 | |||
| US20180365560A1 | 2018-12-20 |
Attorney, Agent or Firm:
TDIP & PARTNERS (CN)
Download PDF: