Parity 1-norm unequal length sequence similarity metric algorithm based on DTW

A sequence similarity and long sequence technology, applied in the field of long sequence data mining algorithms

CN103942444AInactive Publication Date: 2014-07-23衣晓
3 Cites 9 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Publication Date
2014-07-23
Estimated Expiration
Not applicable · inactive patent

Smart Images

  • Figure 1
    Figure 1
  • Figure 2
    Figure 2
  • Figure 3
    Figure 3
Patent Text Reader

Abstract

The invention discloses a parity 1-norm unequal length sequence similarity metric algorithm based on DTW. The parity 1-norm unequal length sequence similarity metric algorithm based on the DTW is oriented to the problem of the similarity metric of unequal length sequences with points of sudden changes. According to the algorithm, an absolute distance matrix of the unequal length sequences is constructed based on the DTW, the minimum values in the minimum absolute distance matrix are extracted in lines or columns to form a minimum absolute distance collection; in order to deal with the influences of the points of the sudden changes on the sequence distance metric, the parity 1-norm of the minimum absolute distance collection is adopted as a distance metric, and lastly, the similarity metric of the unequal length sequences with the points of the sudden changes is formed according to the relation between the distance metric and the similarity.
Need to check novelty before this filing date? Find Prior Art

Description

technical field

[0001] The invention relates to a data fusion algorithm, in particular to a data mining algorithm for unequal-length sequences. Background technique

[0002] As a kind of uncertain data, sequence data is the main research object in the field of data mining, and widely exists in the fields of economic forecasting, medical research, weather forecasting, network security and military science. With the rapid development of information technology, the amount of data is increasing, and the information contained is also increasing, which undoubtedly entered the era of big data. How to mine effective information and knowledge hidden in these data has been extensively studied in recent years. Sequence data is high-dimensional data composed of many data points. The length of these data points may vary over time. Mining sequence data with inconsistent lengths is a key issue in data mining. Sequence similarity measurement method is an important process and basic method...

Examples

Embodiment

[0065] Assuming that two types of sensors sensor1 and sensor2 are set to ESM and ELINT, the target is measured, and three types of target identity information are measured: radar carrier frequency RF, pulse repetition frequency PRF and pulse width PW, in which the sensor receives some kind of electromagnetic interference Make some data in the measurement data deviate from the real value, after the front-end data processing, after correlation, two target sequence matrices to be identified are obtained and , which are composed of three sequences, representing the three types of parameters RF, PRF and PW, and there are mutation points in the sequence due to interference. There are four types of target identity attributes in the target database, respectively using the sequence matrix , , and Indicates that the data parameters in the matrix and the target sequence matrix to be identified and Correspondingly, the lengths between them are unequal.

[0066] Using the DT...