Dynamic classifier chain adjusting method for multi-label classification

A dynamic classification and adjustment method technology, applied in the direction of instruments, character and pattern recognition, computer components, etc., to achieve the effect of alleviating randomness

Inactive Publication Date: 2019-03-19
JINAN INSPUR HIGH TECH TECH DEV CO LTD
View PDF1 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] The technical task of the present invention is to provide a dynamic classifier chain adjustment method for multi-label classification, to solve the problem of how to reduce the probability that the output is misclassified in the eyes near the threshold, and to alleviate the problem of the label prediction order of the classifier chain species. Problems of randomness, uncertainty and instability of classification results

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Dynamic classifier chain adjusting method for multi-label classification

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0027] The dynamic classifier chain adjustment method for multi-label classification of the present invention, the method is in the training data set, respectively counts the number of times each marker and other markers co-occur and sorts them in descending order, and completes the classifier in the training data set Chain sorting, and then randomly select a classifier from the classifier chain to complete the classification of unknown samples. When classifying unknown samples, set two thresholds in advance, according to the output value of the randomly selected classifier and the two thresholds Complete the classification of unknown samples, reduce the probability of misclassification of unknown samples whose output is near the threshold when classifying unknown samples, and alleviate the randomness of the label prediction sequence and the uncertainty of classification results in the classifier chain sex and instability.

Embodiment 2

[0029] The ordering of classifier chains of the present invention refers to: each classifier is a secondary classifier and corresponds to each category, the number of classifier chains is equal to the number of categories, and the first classifier of each classifier chain corresponds to the first classifier. categories.

[0030] The sorting method of the classifier chain includes the following steps:

[0031] (1), mark a category in the training data set as l i ;

[0032] (2), from all have mark l i In the samples, the statistics have the exception mark l i The number of samples marked other than ;

[0033] (3), in descending order, mark l i To sort by tags other than the tag l i Arranged in the first position of the classifier chain, the order of the subsequent classifiers can be sorted according to the marks.

[0034] More preferably, the relationship between the classifier chain and the classifier is specifically as follows: there are n category marks {1 1 , l 2 .....

Embodiment 3

[0036] The classification method of the unknown sample of the present invention specifically includes the following steps:

[0037] (i) When classifying unknown samples, randomly select a classifier C i ;

[0038] (ii), according to C i The output of f(C i ) value and threshold a and threshold b to select the classification of unknown samples:

[0039] ①, when C i When the output value is greater than the threshold b, then according to C i The corresponding classifier chain R(C i ), to classify unknown samples;

[0040] ②, when C i When the output value is less than the threshold a, then from C i The corresponding classifier chain R(C i ) at the end to select the corresponding secondary classifier classification in turn, until a classifier C with output greater than the threshold appears j , then follow C j The corresponding classifier chain R(C j ), to classify unknown samples;

[0041] ③, when C j When the output value of is between a and b, reselect another cl...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention discloses a dynamic classifier chain adjusting method for multi-label classification, belonging to the field of machine learning field. The technical problem to be solved by the invention is how to reduce the probability that the output is wrongly divided in eyes near a threshold value, the randomness of a mark prediction sequence existing in a classifier chain is relieved; uncertainty and instability of classification results are realized; the adopted technical scheme is as follows: the device comprises a base, the method is characterized in that training data are concentrated;respectively counting the co-occurrence frequency of each mark and the marks except the mark, and progressively decreasing and sorting; the method comprises the following steps: selecting a classifierfrom a training data set to complete the sorting of classifier chains in the training data set, randomly selecting one classifier from the classifier chains to complete the classification of unknownsamples, setting two thresholds in advance during the classification of the unknown samples, and completing the classification of the unknown samples according to the output values of the randomly selected classifiers and the sizes of the two thresholds.

Description

technical field [0001] The invention relates to the field of machine learning, in particular to a dynamic classifier chain adjustment method for multi-label classification. Background technique [0002] In traditional binary classification and multi-classification problems, a sample belongs to only one category, but in real life, a sample often belongs to multiple categories at the same time, and has multiple different category labels, so there is a multi-label classification problem. For example, an article about Beckham may belong to both sports news and entertainment news due to its multiple identities; generally the same movie can be in several different movie categories (adventure, action, thriller) Found in , this type of problem is called multi-label or multi-label problem. There is a certain correlation between multiple tags of a sample, such as an action movie, and the probability of being an adventure class is higher than that of a literary class. a lot of. [00...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/62
CPCG06F18/2431
Inventor 郝虹段成德
Owner JINAN INSPUR HIGH TECH TECH DEV CO LTD
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap