Title:
FEATURE SELECTION METHOD AND APPARATUS FOR CONSTRUCTING MACHINE LEARNING MODEL AND DEVICE
Document Type and Number:
WIPO Patent Application WO/2020/082865
Kind Code:
A1
Abstract:
Embodiments of the invention provide a feature selection method and apparatus for constructing a machine learning model and a device. The feature selection method comprises: acquiring a training data set acquired after screening; dividing the training data set according to a pre-configured division method, so as to acquire k training data subsets; concurrently executing the following process on the k training data subsets for a total of k times: selecting k - 1 training data subsets from the k training data subsets as a current training data set, calculating, according to the current training data set, m evaluation indicators of multiple features serving as candidates, sorting the multiple features according to each of the evaluation indicators, so as to acquire m indicator ranking lists of the multiple features, and training a machine learning model on the basis of the current training data set, so as to predict an importance ranking list of the multiple features; fusing the k × m indicator ranking lists and the k importance ranking lists acquired after k times, so as to acquire a final ranking list of the multiple features; and selecting a target feature from the multiple feature according to the final ranking list.
Inventors:
TANG YUZHOU (CN)
JIN HONG (CN)
WANG WEIQIANG (CN)
ZHAO WENBIAO (CN)
JIN HONG (CN)
WANG WEIQIANG (CN)
ZHAO WENBIAO (CN)
Application Number:
PCT/CN2019/101397
Publication Date:
April 30, 2020
Filing Date:
August 19, 2019
Export Citation:
Assignee:
ALIBABA GROUP HOLDING LTD (CN)
International Classes:
G06N20/00
Domestic Patent References:
| WO2018107906A1 | 2018-06-21 |
Foreign References:
| CN109460825A | 2019-03-12 | |||
| CN105825081A | 2016-08-03 | |||
| CN108446741A | 2018-08-24 | |||
| CN104732242A | 2015-06-24 | |||
| US20160358099A1 | 2016-12-08 |
Attorney, Agent or Firm:
BEIJING BESTIPR INTELLECTUAL PROPERTY LAW CORPORATION (CN)
Download PDF: