Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
TIME SERIES DATA TREND FEATURE EXTRACTION METHOD BASED ON DYNAMIC GRID DIVISION
Document Type and Number:
WIPO Patent Application WO/2022/151829
Kind Code:
A1
Abstract:
A time series data trend feature extraction method based on dynamic grid division, the method comprising the following steps: setting a target number N, dynamically dividing grids according to the density distribution of time series data, and respectively dividing the time and value of the time series data into m segments and n segments; traversing local data in each grid, and acquiring a priority queue of key feature points by using a linear segmentation distance calculation means; summarizing the priority queues of the key feature points extracted from the grids, so as to obtain a one-dimensional feature sub-sequence S1 of original data; according to the target number N, extracting data points from a time series at equal intervals to obtain a one-dimensional trend sub-sequence S2 of the original data; and integrating the feature sub-sequence S1 and the trend sub-sequence S2 to obtain a new sequence S for data mining. By means of the method, key feature points and trend information in time series data can be reserved by using a small number of data points, such that the efficiency and accuracy of subsequent data modeling and analysis are improved.

Inventors:
YANG HAOJIE (CN)
YANG YU (CN)
SUN FENGCHENG (CN)
Application Number:
PCT/CN2021/130798
Publication Date:
July 21, 2022
Filing Date:
November 16, 2021
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
HANGZHOU ANMAISHENG INTELLIGENT TECH CO LTD (CN)
International Classes:
G06F17/18
Foreign References:
CN112765562A2021-05-07
CN111143442A2020-05-12
CN108804731A2018-11-13
CN110489810A2019-11-22
CN104820779A2015-08-05
US20190026351A12019-01-24
US20030009399A12003-01-09
CN108804731A2018-11-13
Attorney, Agent or Firm:
HANGZHOU HANGCHENG PATENT ATTORNEY OFFICE CO., LTD. (CN)
Download PDF: